20220616.md

ArXiv cs.CV --Thu, 16 Jun 2022

1.PlanarRecon: Real-time 3D Plane Detection and Reconstruction from Posed Monocular Videos ⬇️

We present PlanarRecon -- a novel framework for globally coherent detection and reconstruction of 3D planes from a posed monocular video. Unlike previous works that detect planes in 2D from a single image, PlanarRecon incrementally detects planes in 3D for each video fragment, which consists of a set of key frames, from a volumetric representation of the scene using neural networks. A learning-based tracking and fusion module is designed to merge planes from previous fragments to form a coherent global plane reconstruction. Such design allows PlanarRecon to integrate observations from multiple views within each fragment and temporal information across different ones, resulting in an accurate and coherent reconstruction of the scene abstraction with low-polygonal geometry. Experiments show that the proposed approach achieves state-of-the-art performances on the ScanNet dataset while being real-time.

2.Variable Bitrate Neural Fields ⬇️

Neural approximations of scalar and vector fields, such as signed distance functions and radiance fields, have emerged as accurate, high-quality representations. State-of-the-art results are obtained by conditioning a neural approximation with a lookup from trainable feature grids that take on part of the learning task and allow for smaller, more efficient neural networks. Unfortunately, these feature grids usually come at the cost of significantly increased memory consumption compared to stand-alone neural network models. We present a dictionary method for compressing such feature grids, reducing their memory consumption by up to 100x and permitting a multiresolution representation which can be useful for out-of-core streaming. We formulate the dictionary optimization as a vector-quantized auto-decoder problem which lets us learn end-to-end discrete neural representations in a space where no direct supervision is available and with dynamic topology and structure. Our source code will be available at this https URL.

3.Masked Frequency Modeling for Self-Supervised Visual Pre-Training ⬇️

We present Masked Frequency Modeling (MFM), a unified frequency-domain-based approach for self-supervised pre-training of visual models. Instead of randomly inserting mask tokens to the input embeddings in the spatial domain, in this paper, we shift the perspective to the frequency domain. Specifically, MFM first masks out a portion of frequency components of the input image and then predicts the missing frequencies on the frequency spectrum. Our key insight is that predicting masked components in the frequency domain is more ideal to reveal underlying image patterns rather than predicting masked patches in the spatial domain, due to the heavy spatial redundancy. Our findings suggest that with the right configuration of mask-and-predict strategy, both the structural information within high-frequency components and the low-level statistics among low-frequency counterparts are useful in learning good representations. For the first time, MFM demonstrates that, for both ViT and CNN, a simple non-Siamese framework can learn meaningful representations even using none of the following: (i) extra data, (ii) extra model, (iii) mask token. Experimental results on ImageNet and several robustness benchmarks show the competitive performance and advanced robustness of MFM compared with recent masked image modeling approaches. Furthermore, we also comprehensively investigate the effectiveness of classical image restoration tasks for representation learning from a unified frequency perspective and reveal their intriguing relations with our MFM approach. Project page: this https URL.

4.LET-3D-AP: Longitudinal Error Tolerant 3D Average Precision for Camera-Only 3D Detection ⬇️

The popular object detection metric 3D Average Precision (3D AP) relies on the intersection over union between predicted bounding boxes and ground truth bounding boxes. However, depth estimation based on cameras has limited accuracy, which may cause otherwise reasonable predictions that suffer from such longitudinal localization errors to be treated as false positives and false negatives. We therefore propose variants of the popular 3D AP metric that are designed to be more permissive with respect to depth estimation errors. Specifically, our novel longitudinal error tolerant metrics, LET-3D-AP and LET-3D-APL, allow longitudinal localization errors of the predicted bounding boxes up to a given tolerance. The proposed metrics have been used in the Waymo Open Dataset 3D Camera-Only Detection Challenge. We believe that they will facilitate advances in the field of camera-only 3D detection by providing more informative performance signals.

5.Waymo Open Dataset: Panoramic Video Panoptic Segmentation ⬇️

Panoptic image segmentation is the computer vision task of finding groups of pixels in an image and assigning semantic classes and object instance identifiers to them. Research in image segmentation has become increasingly popular due to its critical applications in robotics and autonomous driving. The research community thereby relies on publicly available benchmark dataset to advance the state-of-the-art in computer vision. Due to the high costs of densely labeling the images, however, there is a shortage of publicly available ground truth labels that are suitable for panoptic segmentation. The high labeling costs also make it challenging to extend existing datasets to the video domain and to multi-camera setups. We therefore present the Waymo Open Dataset: Panoramic Video Panoptic Segmentation Dataset, a large-scale dataset that offers high-quality panoptic segmentation labels for autonomous driving. We generate our dataset using the publicly available Waymo Open Dataset, leveraging the diverse set of camera images. Our labels are consistent over time for video processing and consistent across multiple cameras mounted on the vehicles for full panoramic scene understanding. Specifically, we offer labels for 28 semantic categories and 2,860 temporal sequences that were captured by five cameras mounted on autonomous vehicles driving in three different geographical locations, leading to a total of 100k labeled camera images. To the best of our knowledge, this makes our dataset an order of magnitude larger than existing datasets that offer video panoptic segmentation labels. We further propose a new benchmark for Panoramic Video Panoptic Segmentation and establish a number of strong baselines based on the DeepLab family of models. We will make the benchmark and the code publicly available. Find the dataset at this https URL.

6.Masked Siamese ConvNets ⬇️

Self-supervised learning has shown superior performances over supervised methods on various vision benchmarks. The siamese network, which encourages embeddings to be invariant to distortions, is one of the most successful self-supervised visual representation learning approaches. Among all the augmentation methods, masking is the most general and straightforward method that has the potential to be applied to all kinds of input and requires the least amount of domain knowledge. However, masked siamese networks require particular inductive bias and practically only work well with Vision Transformers. This work empirically studies the problems behind masked siamese networks with ConvNets. We propose several empirical designs to overcome these problems gradually. Our method performs competitively on low-shot image classification and outperforms previous methods on object detection benchmarks. We discuss several remaining issues and hope this work can provide useful data points for future general-purpose self-supervised learning.

7.Prefix Language Models are Unified Modal Learners ⬇️

With the success of vision-language pre-training, we have witnessed the state-of-the-art has been pushed on multi-modal understanding and generation. However, the current pre-training paradigm is either incapable of targeting all modalities at once (e.g., text generation and image generation), or requires multi-fold well-designed tasks which significantly limits the scalability. We demonstrate that a unified modal model could be learned with a prefix language modeling objective upon text and image sequences. Thanks to the simple but powerful pre-training paradigm, our proposed model, DaVinci, is simple to train, scalable to huge data, and adaptable to a variety of downstream tasks across modalities (language / vision / vision+language), types (understanding / generation) and settings (e.g., zero-shot, fine-tuning, linear evaluation) with a single unified architecture. DaVinci achieves the competitive performance on a wide range of 26 understanding / generation tasks, and outperforms previous unified vision-language models on most tasks, including ImageNet classification (+1.6%), VQAv2 (+1.4%), COCO caption generation (BLEU@4 +1.1%, CIDEr +1.5%) and COCO image generation (IS +0.9%, FID -1.0%), at the comparable model and data scale. Furthermore, we offer a well-defined benchmark for future research by reporting the performance on different scales of the pre-training dataset on a heterogeneous and wide distribution coverage. Our results establish new, stronger baselines for future comparisons at different data scales and shed light on the difficulties of comparing VLP models more generally.

8.Neural Deformable Voxel Grid for Fast Optimization of Dynamic View Synthesis ⬇️

Recently, Neural Radiance Fields (NeRF) is revolutionizing the task of novel view synthesis (NVS) for its superior performance. However, NeRF and its variants generally require a lengthy per-scene training procedure, where a multi-layer perceptron (MLP) is fitted to the captured images. To remedy the challenge, the voxel-grid representation has been proposed to significantly speed up the training. However, these existing methods can only deal with static scenes. How to develop an efficient and accurate dynamic view synthesis method remains an open problem. Extending the methods for static scenes to dynamic scenes is not straightforward as both the scene geometry and appearance change over time. In this paper, built on top of the recent advances in voxel-grid optimization, we propose a fast deformable radiance field method to handle dynamic scenes. Our method consists of two modules. The first module adopts a deformation grid to store 3D dynamic features, and a light-weight MLP for decoding the deformation that maps a 3D point in observation space to the canonical space using the interpolated features. The second module contains a density and a color grid to model the geometry and density of the scene. The occlusion is explicitly modeled to further improve the rendering quality. Experimental results show that our method achieves comparable performance to D-NeRF using only 20 minutes for training, which is more than 70x faster than D-NeRF, clearly demonstrating the efficiency of our proposed method.

9.Diffusion Models for Video Prediction and Infilling ⬇️

To predict and anticipate future outcomes or reason about missing information in a sequence is a key ability for agents to be able to make intelligent decisions. This requires strong temporally coherent generative capabilities. Diffusion models have shown huge success in several generative tasks lately, but have not been extensively explored in the video domain. We present Random-Mask Video Diffusion (RaMViD), which extends image diffusion models to videos using 3D convolutions, and introduces a new conditioning technique during training. By varying the mask we condition on, the model is able to perform video prediction, infilling and upsampling. Since we do not use concatenation to condition on a mask, as done in most conditionally trained diffusion models, we are able to decrease the memory footprint. We evaluated the model on two benchmark datasets for video prediction and one for video generation on which we achieved competitive results. On Kinetics-600 we achieved state-of-the-art for video prediction.

10.VoxGRAF: Fast 3D-Aware Image Synthesis with Sparse Voxel Grids ⬇️

State-of-the-art 3D-aware generative models rely on coordinate-based MLPs to parameterize 3D radiance fields. While demonstrating impressive results, querying an MLP for every sample along each ray leads to slow rendering. Therefore, existing approaches often render low-resolution feature maps and process them with an upsampling network to obtain the final image. Albeit efficient, neural rendering often entangles viewpoint and content such that changing the camera pose results in unwanted changes of geometry or appearance. Motivated by recent results in voxel-based novel view synthesis, we investigate the utility of sparse voxel grid representations for fast and 3D-consistent generative modeling in this paper. Our results demonstrate that monolithic MLPs can indeed be replaced by 3D convolutions when combining sparse voxel grids with progressive growing, free space pruning and appropriate regularization. To obtain a compact representation of the scene and allow for scaling to higher voxel resolutions, our model disentangles the foreground object (modeled in 3D) from the background (modeled in 2D). In contrast to existing approaches, our method requires only a single forward pass to generate a full 3D scene. It hence allows for efficient rendering from arbitrary viewpoints while yielding 3D consistent results with high visual fidelity.

11.A Simple Data Mixing Prior for Improving Self-Supervised Learning ⬇️

Data mixing (e.g., Mixup, Cutmix, ResizeMix) is an essential component for advancing recognition models. In this paper, we focus on studying its effectiveness in the self-supervised setting. By noticing the mixed images that share the same source images are intrinsically related to each other, we hereby propose SDMP, short for $\textbf{S}$imple $\textbf{D}$ata $\textbf{M}$ixing $\textbf{P}$rior, to capture this straightforward yet essential prior, and position such mixed images as additional $\textbf{positive pairs}$ to facilitate self-supervised representation learning. Our experiments verify that the proposed SDMP enables data mixing to help a set of self-supervised learning frameworks (e.g., MoCo) achieve better accuracy and out-of-distribution robustness. More notably, our SDMP is the first method that successfully leverages data mixing to improve (rather than hurt) the performance of Vision Transformers in the self-supervised setting. Code is publicly available at this https URL

12.ELUDE: Generating interpretable explanations via a decomposition into labelled and unlabelled features ⬇️

Deep learning models have achieved remarkable success in different areas of machine learning over the past decade; however, the size and complexity of these models make them difficult to understand. In an effort to make them more interpretable, several recent works focus on explaining parts of a deep neural network through human-interpretable, semantic attributes. However, it may be impossible to completely explain complex models using only semantic attributes. In this work, we propose to augment these attributes with a small set of uninterpretable features. Specifically, we develop a novel explanation framework ELUDE (Explanation via Labelled and Unlabelled DEcomposition) that decomposes a model's prediction into two parts: one that is explainable through a linear combination of the semantic attributes, and another that is dependent on the set of uninterpretable features. By identifying the latter, we are able to analyze the "unexplained" portion of the model, obtaining insights into the information used by the model. We show that the set of unlabelled features can generalize to multiple models trained with the same feature space and compare our work to two popular attribute-oriented methods, Interpretable Basis Decomposition and Concept Bottleneck, and discuss the additional insights ELUDE provides.

13.Structured Video Tokens @ Ego4D PNR Temporal Localization Challenge 2022 ⬇️

This technical report describes the SViT approach for the Ego4D Point of No Return (PNR) Temporal Localization Challenge. We propose a learning framework StructureViT (SViT for short), which demonstrates how utilizing the structure of a small number of images only available during training can improve a video model. SViT relies on two key insights. First, as both images and videos contain structured information, we enrich a transformer model with a set of \emph{object tokens} that can be used across images and videos. Second, the scene representations of individual frames in video should "align" with those of still images. This is achieved via a "Frame-Clip Consistency" loss, which ensures the flow of structured information between images and videos. SViT obtains strong performance on the challenge test set with 0.656 absolute temporal localization error.

14.Residual Sparsity Connection Learning for Efficient Video Super-Resolution ⬇️

Lighter and faster models are crucial for the deployment of video super-resolution (VSR) on resource-limited devices, e.g., smartphones and wearable devices. In this paper, we develop Residual Sparsity Connection Learning (RSCL), a structured pruning scheme, to reduce the redundancy of convolution kernels and obtain a compact VSR network with a minor performance drop. However, residual blocks require the pruned filter indices of skip and residual connections to be the same, which is tricky for pruning. Thus, to mitigate the pruning restrictions of residual blocks, we design a Residual Sparsity Connection (RSC) scheme by preserving the feature channels and only operating on the important channels. Moreover, for the pixel-shuffle operation, we design a special pruning scheme by grouping several filters as pruning units to guarantee the accuracy of feature channel-space conversion after pruning. In addition, we introduce Temporal Finetuning (TF) to reduce the pruning error amplification of hidden states with temporal propagation. Extensive experiments show that the proposed RSCL significantly outperforms recent methods quantitatively and qualitatively. Codes and models will be released.

15.AVATAR: Unconstrained Audiovisual Speech Recognition ⬇️

Audio-visual automatic speech recognition (AV-ASR) is an extension of ASR that incorporates visual cues, often from the movements of a speaker's mouth. Unlike works that simply focus on the lip motion, we investigate the contribution of entire visual frames (visual actions, objects, background etc.). This is particularly useful for unconstrained videos, where the speaker is not necessarily visible. To solve this task, we propose a new sequence-to-sequence AudioVisual ASR TrAnsformeR (AVATAR) which is trained end-to-end from spectrograms and full-frame RGB. To prevent the audio stream from dominating training, we propose different word-masking strategies, thereby encouraging our model to pay attention to the visual stream. We demonstrate the contribution of the visual modality on the How2 AV-ASR benchmark, especially in the presence of simulated noise, and show that our model outperforms all other prior work by a large margin. Finally, we also create a new, real-world test bed for AV-ASR called VisSpeech, which demonstrates the contribution of the visual modality under challenging audio conditions.

16.A Unified Sequence Interface for Vision Tasks ⬇️

While language tasks are naturally expressed in a single, unified, modeling framework, i.e., generating sequences of tokens, this has not been the case in computer vision. As a result, there is a proliferation of distinct architectures and loss functions for different vision tasks. In this work we show that a diverse set of "core" computer vision tasks can also be unified if formulated in terms of a shared pixel-to-sequence interface. We focus on four tasks, namely, object detection, instance segmentation, keypoint detection, and image captioning, all with diverse types of outputs, e.g., bounding boxes or dense masks. Despite that, by formulating the output of each task as a sequence of discrete tokens with a unified interface, we show that one can train a neural network with a single model architecture and loss function on all these tasks, with no task-specific customization. To solve a specific task, we use a short prompt as task description, and the sequence output adapts to the prompt so it can produce task-specific output. We show that such a model can achieve competitive performance compared to well-established task-specific models.

17.SP-ViT: Learning 2D Spatial Priors for Vision Transformers ⬇️

Recently, transformers have shown great potential in image classification and established state-of-the-art results on the ImageNet benchmark. However, compared to CNNs, transformers converge slowly and are prone to overfitting in low-data regimes due to the lack of spatial inductive biases. Such spatial inductive biases can be especially beneficial since the 2D structure of an input image is not well preserved in transformers. In this work, we present Spatial Prior-enhanced Self-Attention (SP-SA), a novel variant of vanilla Self-Attention (SA) tailored for vision transformers. Spatial Priors (SPs) are our proposed family of inductive biases that highlight certain groups of spatial relations. Unlike convolutional inductive biases, which are forced to focus exclusively on hard-coded local regions, our proposed SPs are learned by the model itself and take a variety of spatial relations into account. Specifically, the attention score is calculated with emphasis on certain kinds of spatial relations at each head, and such learned spatial foci can be complementary to each other. Based on SP-SA we propose the SP-ViT family, which consistently outperforms other ViT models with similar GFlops or parameters. Our largest model SP-ViT-L achieves a record-breaking 86.3% Top-1 accuracy with a reduction in the number of parameters by almost 50% compared to previous state-of-the-art model (150M for SP-ViT-L vs 271M for CaiT-M-36) among all ImageNet-1K models trained on 224x224 and fine-tuned on 384x384 resolution w/o extra data.

18.Coarse-to-Fine Vision-Language Pre-training with Fusion in the Backbone ⬇️

Vision-language (VL) pre-training has recently received considerable attention. However, most existing end-to-end pre-training approaches either only aim to tackle VL tasks such as image-text retrieval, visual question answering (VQA) and image captioning that test high-level understanding of images, or only target region-level understanding for tasks such as phrase grounding and object detection. We present FIBER (Fusion-In-the-Backbone-based transformER), a new VL model architecture that can seamlessly handle both these types of tasks. Instead of having dedicated transformer layers for fusion after the uni-modal backbones, FIBER pushes multimodal fusion deep into the model by inserting cross-attention into the image and text backbones, bringing gains in terms of memory and performance. In addition, unlike previous work that is either only pre-trained on image-text data or on fine-grained data with box-level annotations, we present a two-stage pre-training strategy that uses both these kinds of data efficiently: (i) coarse-grained pre-training based on image-text data; followed by (ii) fine-grained pre-training based on image-text-box data. We conduct comprehensive experiments on a wide range of VL tasks, ranging from VQA, image captioning, and retrieval, to phrase grounding, referring expression comprehension, and object detection. Using deep multimodal fusion coupled with the two-stage pre-training, FIBER provides consistent performance improvements over strong baselines across all tasks, often outperforming methods using magnitudes more data. Code is available at this https URL.

19.Real3D-Aug: Point Cloud Augmentation by Placing Real Objects with Occlusion Handling for 3D Detection and Segmentation ⬇️

Object detection and semantic segmentation with the 3D lidar point cloud data require expensive annotation. We propose a data augmentation method that takes advantage of already annotated data multiple times. We propose an augmentation framework that reuses real data, automatically finds suitable placements in the scene to be augmented, and handles occlusions explicitly. Due to the usage of the real data, the scan points of newly inserted objects in augmentation sustain the physical characteristics of the lidar, such as intensity and raydrop. The pipeline proves competitive in training top-performing models for 3D object detection and semantic segmentation. The new augmentation provides a significant performance gain in rare and essential classes, notably 6.65% average precision gain for "Hard" pedestrian class in KITTI object detection or 2.14 mean IoU gain in the SemanticKITTI segmentation challenge over the state of the art.

20.Evaluating object detector ensembles for improving the robustness of artifact detection in endoscopic video streams ⬇️

In this contribution we use an ensemble deep-learning method for combining the prediction of two individual one-stage detectors (i.e., YOLOv4 and Yolact) with the aim to detect artefacts in endoscopic images. This ensemble strategy enabled us to improve the robustness of the individual models without harming their real-time computation capabilities. We demonstrated the effectiveness of our approach by training and testing the two individual models and various ensemble configurations on the "Endoscopic Artifact Detection Challenge" dataset. Extensive experiments show the superiority, in terms of mean average precision, of the ensemble approach over the individual models and previous works in the state of the art.

21.A Meta-Analysis of Distributionally-Robust Models ⬇️

State-of-the-art image classifiers trained on massive datasets (such as ImageNet) have been shown to be vulnerable to a range of both intentional and incidental distribution shifts. On the other hand, several recent classifiers with favorable out-of-distribution (OOD) robustness properties have emerged, achieving high accuracy on their target tasks while maintaining their in-distribution accuracy on challenging benchmarks. We present a meta-analysis on a wide range of publicly released models, most of which have been published over the last twelve months. Through this meta-analysis, we empirically identify four main commonalities for all the best-performing OOD-robust models, all of which illuminate the considerable promise of vision-language pre-training.

22.How to Reduce Change Detection to Semantic Segmentation ⬇️

Change detection (CD) aims to identify changes that occur in an image pair taken different times. Prior methods devise specific networks from scratch to predict change masks in pixel-level, and struggle with general segmentation problems. In this paper, we propose a new paradigm that reduces CD to semantic segmentation which means tailoring an existing and powerful semantic segmentation network to solve CD. This new paradigm conveniently enjoys the mainstream semantic segmentation techniques to deal with general segmentation problems in CD. Hence we can concentrate on studying how to detect changes. We propose a novel and importance insight that different change types exist in CD and they should be learned separately. Based on it, we devise a module named MTF to extract the change information and fuse temporal features. MTF enjoys high interpretability and reveals the essential characteristic of CD. And most segmentation networks can be adapted to solve the CD problems with our MTF module. Finally, we propose C-3PO, a network to detect changes at pixel-level. C-3PO achieves state-of-the-art performance without bells and whistles. It is simple but effective and can be considered as a new baseline in this field. Our code will be available.

23.Deep Multi-Task Networks For Occluded Pedestrian Pose Estimation ⬇️

Most of the existing works on pedestrian pose estimation do not consider estimating the pose of an occluded pedestrians, as the annotations of the occluded parts are not available in relevant automotive datasets. For example, CityPersons, a well-known dataset for pedestrian detection in automotive scenes does not provide pose annotations, whereas MS-COCO, a non-automotive dataset, contains human pose estimation. In this work, we propose a multi-task framework to extract pedestrian features through detection and instance segmentation tasks performed separately on these two distributions. Thereafter, an encoder learns pose specific features using an unsupervised instance-level domain adaptation method for the pedestrian instances from both distributions. The proposed framework has improved state-of-the-art performances of pose estimation, pedestrian detection, and instance segmentation.

24.PolyU-BPCoMa: A Dataset and Benchmark Towards Mobile Colorized Mapping Using a Backpack Multisensorial System ⬇️

Constructing colorized point clouds from mobile laser scanning and images is a fundamental work in surveying and mapping. It is also an essential prerequisite for building digital twins for smart cities. However, existing public datasets are either in relatively small scales or lack accurate geometrical and color ground truth. This paper documents a multisensorial dataset named PolyU-BPCoMA which is distinctively positioned towards mobile colorized mapping. The dataset incorporates resources of 3D LiDAR, spherical imaging, GNSS and IMU on a backpack platform. Color checker boards are pasted in each surveyed area as targets and ground truth data are collected by an advanced terrestrial laser scanner (TLS). 3D geometrical and color information can be recovered in the colorized point clouds produced by the backpack system and the TLS, respectively. Accordingly, we provide an opportunity to benchmark the mapping and colorization accuracy simultaneously for a mobile multisensorial system. The dataset is approximately 800 GB in size covering both indoor and outdoor environments. The dataset and development kits are available at this https URL.

25.Coarse-to-fine Deep Video Coding with Hyperprior-guided Mode Prediction ⬇️

The previous deep video compression approaches only use the single scale motion compensation strategy and rarely adopt the mode prediction technique from the traditional standards like H.264/H.265 for both motion and residual compression. In this work, we first propose a coarse-to-fine (C2F) deep video compression framework for better motion compensation, in which we perform motion estimation, compression and compensation twice in a coarse to fine manner. Our C2F framework can achieve better motion compensation results without significantly increasing bit costs. Observing hyperprior information (i.e., the mean and variance values) from the hyperprior networks contains discriminant statistical information of different patches, we also propose two efficient hyperprior-guided mode prediction methods. Specifically, using hyperprior information as the input, we propose two mode prediction networks to respectively predict the optimal block resolutions for better motion coding and decide whether to skip residual information from each block for better residual coding without introducing additional bit cost while bringing negligible extra computation cost. Comprehensive experimental results demonstrate our proposed C2F video compression framework equipped with the new hyperprior-guided mode prediction methods achieves the state-of-the-art performance on HEVC, UVG and MCL-JCV datasets.

26.READ: Aggregating Reconstruction Error into Out-of-distribution Detection ⬇️

Detecting out-of-distribution (OOD) samples is crucial to the safe deployment of a classifier in the real world. However, deep neural networks are known to be overconfident for abnormal data. Existing works directly design score function by mining the inconsistency from classifier for in-distribution (ID) and OOD. In this paper, we further complement this inconsistency with reconstruction error, based on the assumption that an autoencoder trained on ID data can not reconstruct OOD as well as ID. We propose a novel method, READ (Reconstruction Error Aggregated Detector), to unify inconsistencies from classifier and autoencoder. Specifically, the reconstruction error of raw pixels is transformed to latent space of classifier. We show that the transformed reconstruction error bridges the semantic gap and inherits detection performance from the original. Moreover, we propose an adjustment strategy to alleviate the overconfidence problem of autoencoder according to a fine-grained characterization of OOD data. Under two scenarios of pre-training and retraining, we respectively present two variants of our method, namely READ-MD (Mahalanobis Distance) only based on pre-trained classifier and READ-ED (Euclidean Distance) which retrains the classifier. Our methods do not require access to test time OOD data for fine-tuning hyperparameters. Finally, we demonstrate the effectiveness of the proposed methods through extensive comparisons with state-of-the-art OOD detection algorithms. On a CIFAR-10 pre-trained WideResNet, our method reduces the average FPR@95TPR by up to 9.8% compared with previous state-of-the-art.

27.VisageSynTalk: Unseen Speaker Video-to-Speech Synthesis via Speech-Visage Feature Selection ⬇️

The goal of this work is to reconstruct speech from a silent talking face video. Recent studies have shown impressive performance on synthesizing speech from silent talking face videos. However, they have not explicitly considered on varying identity characteristics of different speakers, which place a challenge in the video-to-speech synthesis, and this becomes more critical in unseen-speaker settings. Distinct from the previous methods, our approach is to separate the speech content and the visage-style from a given silent talking face video. By guiding the model to independently focus on modeling the two representations, we can obtain the speech of high intelligibility from the model even when the input video of an unseen subject is given. To this end, we introduce speech-visage selection module that separates the speech content and the speaker identity from the visual features of the input video. The disentangled representations are jointly incorporated to synthesize speech through visage-style based synthesizer which generates speech by coating the visage-styles while maintaining the speech content. Thus, the proposed framework brings the advantage of synthesizing the speech containing the right content even when the silent talking face video of an unseen subject is given. We validate the effectiveness of the proposed framework on the GRID, TCD-TIMIT volunteer, and LRW datasets. The synthesized speech can be heard in supplementary materials.

28.Forecasting of depth and ego-motion with transformers and self-supervision ⬇️

This paper addresses the problem of end-to-end self-supervised forecasting of depth and ego motion. Given a sequence of raw images, the aim is to forecast both the geometry and ego-motion using a self supervised photometric loss. The architecture is designed using both convolution and transformer modules. This leverages the benefits of both modules: Inductive bias of CNN, and the multi-head attention of transformers, thus enabling a rich spatio-temporal representation that enables accurate depth forecasting. Prior work attempts to solve this problem using multi-modal input/output with supervised ground-truth data which is not practical since a large annotated dataset is required. Alternatively to prior methods, this paper forecasts depth and ego motion using only self-supervised raw images as input. The approach performs significantly well on the KITTI dataset benchmark with several performance criteria being even comparable to prior non-forecasting self-supervised monocular depth inference methods.

29.Self-Supervised Implicit Attention: Guided Attention by The Model Itself ⬇️

We propose Self-Supervised Implicit Attention (SSIA), a new approach that adaptively guides deep neural network models to gain attention by exploiting the properties of the models themselves. SSIA is a novel attention mechanism that does not require any extra parameters, computation, or memory access costs during inference, which is in contrast to existing attention mechanism. In short, by considering attention weights as higher-level semantic information, we reconsidered the implementation of existing attention mechanisms and further propose generating supervisory signals from higher network layers to guide lower network layers for parameter updates. We achieved this by building a self-supervised learning task using the hierarchical features of the network itself, which only works at the training stage. To verify the effectiveness of SSIA, we performed a particular implementation (called an SSIA block) in convolutional neural network models and validated it on several image classification datasets. The experimental results show that an SSIA block can significantly improve the model performance, even outperforms many popular attention methods that require additional parameters and computation costs, such as Squeeze-and-Excitation and Convolutional Block Attention Module. Our implementation will be available on GitHub.

30.Physically-admissible polarimetric data augmentation for road-scene analysis ⬇️

Polarimetric imaging, along with deep learning, has shown improved performances on different tasks including scene analysis. However, its robustness may be questioned because of the small size of the training datasets. Though the issue could be solved by data augmentation, polarization modalities are subject to physical feasibility constraints unaddressed by classical data augmentation techniques. To address this issue, we propose to use CycleGAN, an image translation technique based on deep generative models that solely relies on unpaired data, to transfer large labeled road scene datasets to the polarimetric domain. We design several auxiliary loss terms that, alongside the CycleGAN losses, deal with the physical constraints of polarimetric images. The efficiency of this solution is demonstrated on road scene object detection tasks where generated realistic polarimetric images allow to improve performances on cars and pedestrian detection up to 9%. The resulting constrained CycleGAN is publicly released, allowing anyone to generate their own polarimetric images.

31.Zero-shot object goal visual navigation ⬇️

Object goal visual navigation is a challenging task that aims to guide a robot to find the target object only based on its visual observation, and the target is limited to the classes specified in the training stage. However, in real households, there may exist numerous object classes that the robot needs to deal with, and it is hard for all of these classes to be contained in the training stage. To address this challenge, we propose a zero-shot object navigation task by combining zero-shot learning with object goal visual navigation, which aims at guiding robots to find objects belonging to novel classes without any training samples. This task gives rise to the need to generalize the learned policy to novel classes, which is a less addressed issue of object navigation using deep reinforcement learning. To address this issue, we utilize "class-unrelated" data as input to alleviate the overfitting of the classes specified in the training stage. The class-unrelated input consists of detection results and cosine similarity of word embeddings, and does not contain any class-related visual features or knowledge graphs. Extensive experiments on the AI2-THOR platform show that our model outperforms the baseline models in both seen and unseen classes, which proves that our model is less class-sensitive and generalizes better. Our code is available at this https URL

32.Efficient Adaptive Ensembling for Image Classification ⬇️

In recent times, except for sporadic cases, the trend in Computer Vision is to achieve minor improvements over considerable increases in complexity.
To reverse this tendency, we propose a novel method to boost image classification performances without an increase in complexity.
To this end, we revisited ensembling, a powerful approach, not often adequately used due to its nature of increased complexity and training time, making it viable by specific design choices. First, we trained end-to-end two EfficientNet-b0 models (known to be the architecture with the best overall accuracy/complexity trade-off in image classification) on disjoint subsets of data (i.e. bagging). Then, we made an efficient adaptive ensemble by performing fine-tuning of a trainable combination layer. In this way, we were able to outperform the state-of-the-art by an average of 0.5% on the accuracy with restrained complexity both in terms of number of parameters (by 5-60 times), and FLoating point Operations Per Second (by 10-100 times) on several major benchmark datasets, fully embracing the green AI.

33.Ultra Fast Deep Lane Detection with Hybrid Anchor Driven Ordinal Classification ⬇️

Modern methods mainly regard lane detection as a problem of pixel-wise segmentation, which is struggling to address the problems of efficiency and challenging scenarios like severe occlusions and extreme lighting conditions. Inspired by human perception, the recognition of lanes under severe occlusions and extreme lighting conditions is mainly based on contextual and global information. Motivated by this observation, we propose a novel, simple, yet effective formulation aiming at ultra fast speed and the problem of challenging scenarios. Specifically, we treat the process of lane detection as an anchor-driven ordinal classification problem using global features. First, we represent lanes with sparse coordinates on a series of hybrid (row and column) anchors. With the help of the anchor-driven representation, we then reformulate the lane detection task as an ordinal classification problem to get the coordinates of lanes. Our method could significantly reduce the computational cost with the anchor-driven representation. Using the large receptive field property of the ordinal classification formulation, we could also handle challenging scenarios. Extensive experiments on four lane detection datasets show that our method could achieve state-of-the-art performance in terms of both speed and accuracy. A lightweight version could even achieve 300+ frames per second(FPS). Our code is at this https URL.

34.MonoGround: Detecting Monocular 3D Objects from the Ground ⬇️

Monocular 3D object detection has attracted great attention for its advantages in simplicity and cost. Due to the ill-posed 2D to 3D mapping essence from the monocular imaging process, monocular 3D object detection suffers from inaccurate depth estimation and thus has poor 3D detection results. To alleviate this problem, we propose to introduce the ground plane as a prior in the monocular 3d object detection. The ground plane prior serves as an additional geometric condition to the ill-posed mapping and an extra source in depth estimation. In this way, we can get a more accurate depth estimation from the ground. Meanwhile, to take full advantage of the ground plane prior, we propose a depth-align training strategy and a precise two-stage depth inference method tailored for the ground plane prior. It is worth noting that the introduced ground plane prior requires no extra data sources like LiDAR, stereo images, and depth information. Extensive experiments on the KITTI benchmark show that our method could achieve state-of-the-art results compared with other methods while maintaining a very fast speed. Our code and models are available at this https URL.

35.XMorpher: Full Transformer for Deformable Medical Image Registration via Cross Attention ⬇️

An effective backbone network is important to deep learning-based Deformable Medical Image Registration (DMIR), because it extracts and matches the features between two images to discover the mutual correspondence for fine registration. However, the existing deep networks focus on single image situation and are limited in registration task which is performed on paired images. Therefore, we advance a novel backbone network, XMorpher, for the effective corresponding feature representation in DMIR. 1) It proposes a novel full transformer architecture including dual parallel feature extraction networks which exchange information through cross attention, thus discovering multi-level semantic correspondence while extracting respective features gradually for final effective registration. 2) It advances the Cross Attention Transformer (CAT) blocks to establish the attention mechanism between images which is able to find the correspondence automatically and prompts the features to fuse efficiently in the network. 3) It constrains the attention computation between base windows and searching windows with different sizes, and thus focuses on the local transformation of deformable registration and enhances the computing efficiency at the same time. Without any bells and whistles, our XMorpher gives Voxelmorph 2.8% improvement on DSC , demonstrating its effective representation of the features from the paired images in DMIR. We believe that our XMorpher has great application potential in more paired medical images. Our XMorpher is open on this https URL

36.Unsupervised Capsule Networks of High-Dimension Point Clouds classification ⬇️

Three-dimensional point clouds learning is widely applied, but the point clouds are still unable to deal with classification and recognition tasks satisfactorily in the cases of irregular geometric structures and high-dimensional space. In 3D space, point clouds tend to have regular Euclidean structure because of their density. On the contrary, due to the high dimensionality, the spatial structure of high-dimensional space is more complex, and point clouds are mostly presented in non-European structure. Furthermore, among current 3D point clouds classification algorithms, Canonical Capsules algorithm based on Euclidean distance is difficult to decompose and identify non-Euclidean structures effectively. Thus, aiming at the point clouds classification task of non-Euclidean structure in 3D and high-dimensional space, this paper refers to the LLE algorithm based on geodesic distance for optimizing and proposes the unsupervised algorithm of high-dimensional point clouds capsule. In this paper, the geometric features of point clouds are considered in the extraction process, so as to transform the high-dimensional non-Euclidean structure into a lower-dimensional Euclidean structure with retaining spatial geometric features. To verify the feasibility of the unsupervised algorithm of high-dimensional point clouds capsule, experiments are conducted in Swiss Roll dataset, point clouds MNIST dataset and point clouds LFW dataset. The results show that (1) non-Euclidean structures can be can effectively identified by this model in Swiss Roll dataset; (2) a significant unsupervised learning effect is realized in point clouds MNIST dataset. In conclusion, the high-dimensional point clouds capsule unsupervised algorithm proposed in this paper is conducive to expand the application scenarios of current point clouds classification and recognition tasks.

37.Automatic Detection of Rice Disease in Images of Various Leaf Sizes ⬇️

Fast, accurate and affordable rice disease detection method is required to assist rice farmers tackling equipment and expertise shortages problems. In this paper, we focused on the solution using computer vision technique to detect rice diseases from rice field photograph images. Dealing with images took in real-usage situation by general farmers is quite challenging due to various environmental factors, and rice leaf object size variation is one major factor caused performance gradation. To solve this problem, we presented a technique combining a CNN object detection with image tiling technique, based on automatically estimated width size of rice leaves in the images as a size reference for dividing the original input image. A model to estimate leaf width was created by small size CNN such as 18 layer ResNet architecture model. A new divided tiled sub-image set with uniformly sized object was generated and used as input for training a rice disease prediction model. Our technique was evaluated on 4,960 images of eight different types of rice leaf diseases, including blast, blight, brown spot, narrow brown spot, orange, red stripe, rice grassy stunt virus, and streak disease. The mean absolute percentage error (MAPE) for leaf width prediction task evaluated on all eight classes was 11.18% in the experiment, indicating that the leaf width prediction model performed well. The mean average precision (mAP) of the prediction performance on YOLOv4 architecture was enhanced from 87.56% to 91.14% when trained and tested with the tiled dataset. According to our study, the proposed image tiling technique improved rice disease detection efficiency.

38.Recent Advances in Scene Image Representation and Classification ⬇️

With the rise of deep learning algorithms nowadays, scene image representation methods on big data (e.g., SUN-397) have achieved a significant performance boost in classification. However, the performance is still limited because the scene images are mostly complex in nature having higher intra-class dissimilarity and inter-class similarity problems. To deal with such problems, there are several methods proposed in the literature with their own advantages and limitations. A detailed study of previous works is necessary to understand their pros and cons in image representation and classification. In this paper, we review the existing scene image representation methods that are being used widely for image classification. For this, we, first, devise the taxonomy using the seminal existing methods proposed in the literature to this date. Next, we compare their performance both qualitatively (e.g., quality of outputs, pros/cons, etc.) and quantitatively (e.g., accuracy). Last, we speculate the prominent research directions in scene image representation tasks. Overall, this survey provides in-depth insights and applications of recent scene image representation methods for traditional Computer Vision (CV)-based methods, Deep Learning (DL)-based methods, and Search Engine (SE)-based methods.

39.VCT: A Video Compression Transformer ⬇️

We show how transformers can be used to vastly simplify neural video compression. Previous methods have been relying on an increasing number of architectural biases and priors, including motion prediction and warping operations, resulting in complex models. Instead, we independently map input frames to representations and use a transformer to model their dependencies, letting it predict the distribution of future representations given the past. The resulting video compression transformer outperforms previous methods on standard video compression data sets. Experiments on synthetic data show that our model learns to handle complex motion patterns such as panning, blurring and fading purely from data. Our approach is easy to implement, and we release code to facilitate future research.

40.S$^2$-FPN: Scale-ware Strip Attention Guided Feature Pyramid Network for Real-time Semantic Segmentation ⬇️

Modern high-performance semantic segmentation methods employ a heavy backbone and dilated convolution to extract the relevant feature. Although extracting features with both contextual and semantic information is critical for the segmentation tasks, it brings a memory footprint and high computation cost for real-time applications. This paper presents a new model to achieve a trade-off between accuracy/speed for real-time road scene semantic segmentation. Specifically, we proposed a lightweight model named Scale-aware Strip Attention Guided Feature Pyramid Network (S$^2$-FPN). Our network consists of three main modules: Attention Pyramid Fusion (APF) module, Scale-aware Strip Attention Module (SSAM), and Global Feature Upsample (GFU) module. APF adopts an attention mechanisms to learn discriminative multi-scale features and help close the semantic gap between different levels. APF uses the scale-aware attention to encode global context with vertical stripping operation and models the long-range dependencies, which helps relate pixels with similar semantic label. In addition, APF employs channel-wise reweighting block (CRB) to emphasize the channel features. Finally, the decoder of S$^2$-FPN then adopts GFU, which is used to fuse features from APF and the encoder. Extensive experiments have been conducted on two challenging semantic segmentation benchmarks, which demonstrate that our approach achieves better accuracy/speed trade-off with different model settings. The proposed models have achieved a results of 76.2%mIoU/87.3FPS, 77.4%mIoU/67FPS, and 77.8%mIoU/30.5FPS on Cityscapes dataset, and 69.6%mIoU,71.0% mIoU, and 74.2% mIoU on Camvid dataset. The code for this work will be made available at \url{this https URL

41.Human Eyes Inspired Recurrent Neural Networks are More Robust Against Adversarial Noises ⬇️

Compared to human vision, computer vision based on convolutional neural networks (CNN) are more vulnerable to adversarial noises. This difference is likely attributable to how the eyes sample visual input and how the brain processes retinal samples through its dorsal and ventral visual pathways, which are under-explored for computer vision. Inspired by the brain, we design recurrent neural networks, including an input sampler that mimics the human retina, a dorsal network that guides where to look next, and a ventral network that represents the retinal samples. Taking these modules together, the models learn to take multiple glances at an image, attend to a salient part at each glance, and accumulate the representation over time to recognize the image. We test such models for their robustness against a varying level of adversarial noises with a special focus on the effect of different input sampling strategies. Our findings suggest that retinal foveation and sampling renders a model more robust against adversarial noises, and the model may correct itself from an attack when it is given a longer time to take more glances at an image. In conclusion, robust visual recognition can benefit from the combined use of three brain-inspired mechanisms: retinal transformation, attention guided eye movement, and recurrent processing, as opposed to feedforward-only CNNs.

42.Machine vision for vial positioning detection toward the safe automation of material synthesis ⬇️

Although robot-based automation in chemistry laboratories can accelerate the material development process, surveillance-free environments may lead to dangerous accidents primarily due to machine control errors. Object detection techniques can play vital roles in addressing these safety issues; however, state-of-the-art detectors, including single-shot detector (SSD) models, suffer from insufficient accuracy in environments involving complex and noisy scenes. With the aim of improving safety in a surveillance-free laboratory, we report a novel deep learning (DL)-based object detector, namely, DenseSSD. For the foremost and frequent problem of detecting vial positions, DenseSSD achieved a mean average precision (mAP) over 95% based on a complex dataset involving both empty and solution-filled vials, greatly exceeding those of conventional detectors; such high precision is critical to minimizing failure-induced accidents. Additionally, DenseSSD was observed to be highly insensitive to the environmental changes, maintaining its high precision under the variations of solution colors or testing view angles. The robustness of DenseSSD would allow the utilized equipment settings to be more flexible. This work demonstrates that DenseSSD is useful for enhancing safety in an automated material synthesis environment, and it can be extended to various applications where high detection accuracy and speed are both needed.

43.Rethinking Generalization in Few-Shot Classification ⬇️

Single image-level annotations only correctly describe an often small subset of an image's content, particularly when complex real-world scenes are depicted. While this might be acceptable in many classification scenarios, it poses a significant challenge for applications where the set of classes differs significantly between training and test time. In this paper, we take a closer look at the implications in the context of $\textit{few-shot learning}$. Splitting the input samples into patches and encoding these via the help of Vision Transformers allows us to establish semantic correspondences between local regions across images and independent of their respective class. The most informative patch embeddings for the task at hand are then determined as a function of the support set via online optimization at inference time, additionally providing visual interpretability of `$\textit{what matters most}$' in the image. We build on recent advances in unsupervised training of networks via masked image modelling to overcome the lack of fine-grained labels and learn the more general statistical structure of the data while avoiding negative image-level annotation influence, $\textit{aka}$ supervision collapse. Experimental results show the competitiveness of our approach, achieving new state-of-the-art results on four popular few-shot classification benchmarks for $5$-shot and $1$-shot scenarios.

44.Self-Supervised Learning of Image Scale and Orientation ⬇️

We study the problem of learning to assign a characteristic pose, i.e., scale and orientation, for an image region of interest. Despite its apparent simplicity, the problem is non-trivial; it is hard to obtain a large-scale set of image regions with explicit pose annotations that a model directly learns from. To tackle the issue, we propose a self-supervised learning framework with a histogram alignment technique. It generates pairs of image patches by random rescaling/rotating and then train an estimator to predict their scale/orientation values so that their relative difference is consistent with the rescaling/rotating used. The estimator learns to predict a non-parametric histogram distribution of scale/orientation without any supervision. Experiments show that it significantly outperforms previous methods in scale/orientation estimation and also improves image matching and 6 DoF camera pose estimation by incorporating our patch poses into a matching process.

45.GRAM-HD: 3D-Consistent Image Generation at High Resolution with Generative Radiance Manifolds ⬇️

Recent works have shown that 3D-aware GANs trained on unstructured single image collections can generate multiview images of novel instances. The key underpinnings to achieve this are a 3D radiance field generator and a volume rendering process. However, existing methods either cannot generate high-resolution images (e.g., up to 256X256) due to the high computation cost of neural volume rendering, or rely on 2D CNNs for image-space upsampling which jeopardizes the 3D consistency across different views. This paper proposes a novel 3D-aware GAN that can generate high resolution images (up to 1024X1024) while keeping strict 3D consistency as in volume rendering. Our motivation is to achieve super-resolution directly in the 3D space to preserve 3D consistency. We avoid the otherwise prohibitively-expensive computation cost by applying 2D convolutions on a set of 2D radiance manifolds defined in the recent generative radiance manifold (GRAM) approach, and apply dedicated loss functions for effective GAN training at high resolution. Experiments on FFHQ and AFHQv2 datasets show that our method can produce high-quality 3D-consistent results that significantly outperform existing methods.

46.Test-Time Adaptation for Visual Document Understanding ⬇️

Self-supervised pretraining has been able to produce transferable representations for various visual document understanding (VDU) tasks. However, the ability of such representations to adapt to new distribution shifts at test-time has not been studied yet. We propose DocTTA, a novel test-time adaptation approach for documents that leverages cross-modality self-supervised learning via masked visual language modeling as well as pseudo labeling to adapt models learned on a \textit{source} domain to an unlabeled \textit{target} domain at test time. We also introduce new benchmarks using existing public datasets for various VDU tasks including entity recognition, key-value extraction, and document visual question answering tasks where DocTTA improves the source model performance up to 1.79% in (F1 score), 3.43% (F1 score), and 17.68% (ANLS score), respectively while drastically reducing calibration error on target data.

47.Multimodal Event Graphs: Towards Event Centric Understanding of Multimodal World ⬇️

Understanding how events described or shown in multimedia content relate to one another is a critical component to developing robust artificially intelligent systems which can reason about real-world media. While much research has been devoted to event understanding in the text, image, and video domains, none have explored the complex relations that events experience across domains. For example, a news article may describe a protest' event while a video shows an arrest' event. Recognizing that the visual arrest' event is a subevent of the broader protest' event is a challenging, yet important problem that prior work has not explored. In this paper, we propose the novel task of MultiModal Event Event Relations to recognize such cross-modal event relations. We contribute a large-scale dataset consisting of 100k video-news article pairs, as well as a benchmark of densely annotated data. We also propose a weakly supervised multimodal method which integrates commonsense knowledge from an external knowledge base (KB) to predict rich multimodal event hierarchies. Experiments show that our model outperforms a number of competitive baselines on our proposed benchmark. We also perform a detailed analysis of our model's performance and suggest directions for future research.

48.Surgical Phase Recognition in Laparoscopic Cholecystectomy ⬇️

Automatic recognition of surgical phases in surgical videos is a fundamental task in surgical workflow analysis. In this report, we propose a Transformer-based method that utilizes calibrated confidence scores for a 2-stage inference pipeline, which dynamically switches between a baseline model and a separately trained transition model depending on the calibrated confidence level. Our method outperforms the baseline model on the Cholec80 dataset, and can be applied to a variety of action segmentation methods.

49.Automated image analysis in large-scale cellular electron microscopy: A literature survey ⬇️

Large-scale electron microscopy (EM) datasets generated using (semi-) automated microscopes are becoming the standard in EM. Given the vast amounts of data, manual analysis of all data is not feasible, thus automated analysis is crucial. The main challenges in automated analysis include the annotation that is needed to analyse and interpret biomedical images, coupled with achieving high-throughput. Here, we review the current state-of-the-art of automated computer techniques and major challenges for the analysis of structures in cellular EM. The advanced computer vision, deep learning and software tools that have been developed in the last five years for automatic biomedical image analysis are discussed with respect to annotation, segmentation and scalability for EM data. Integration of automatic image acquisition and analysis will allow for high-throughput analysis of millimeter-range datasets with nanometer resolution.

50.DeepRecon: Joint 2D Cardiac Segmentation and 3D Volume Reconstruction via A Structure-Specific Generative Method ⬇️

Joint 2D cardiac segmentation and 3D volume reconstruction are fundamental to building statistical cardiac anatomy models and understanding functional mechanisms from motion patterns. However, due to the low through-plane resolution of cine MR and high inter-subject variance, accurately segmenting cardiac images and reconstructing the 3D volume are challenging. In this study, we propose an end-to-end latent-space-based framework, DeepRecon, that generates multiple clinically essential outcomes, including accurate image segmentation, synthetic high-resolution 3D image, and 3D reconstructed volume. Our method identifies the optimal latent representation of the cine image that contains accurate semantic information for cardiac structures. In particular, our model jointly generates synthetic images with accurate semantic information and segmentation of the cardiac structures using the optimal latent representation. We further explore downstream applications of 3D shape reconstruction and 4D motion pattern adaptation by the different latent-space manipulation strategies.The simultaneously generated high-resolution images present a high interpretable value to assess the cardiac shape and motion.Experimental results demonstrate the effectiveness of our approach on multiple fronts including 2D segmentation, 3D reconstruction, downstream 4D motion pattern adaption performance.

51.Category-Agnostic 6D Pose Estimation with Conditional Neural Processes ⬇️

We present a novel meta-learning approach for 6D pose estimation on unknown objects. In contrast to "instance-level" pose estimation methods, our algorithm learns object representation in a category-agnostic way, which endows it with strong generalization capabilities within and across object categories. Specifically, we employ a conditional neural process-based meta-learning approach to train an encoder to capture texture and geometry of an object in a latent representation, based on very few RGB-D images and ground-truth keypoints. The latent representation is then used by a simultaneously meta-trained decoder to predict the 6D pose of the object in new images. To evaluate our algorithm, experiments are conducted on our new fully-annotated synthetic datasets generated from Multiple Categories in Multiple Scenes (MCMS). Experimental results demonstrate that our model performs well on unseen objects with various shapes and appearances.

52.LAVENDER: Unifying Video-Language Understanding as Masked Language Modeling ⬇️

Unified vision-language frameworks have greatly advanced in recent years, most of which adopt an encoder-decoder architecture to unify image-text tasks as sequence-to-sequence generation. However, existing video-language (VidL) models still require task-specific designs in model architecture and training objectives for each task. In this work, we explore a unified VidL framework LAVENDER, where Masked Language Modeling (MLM) is used as the common interface for all pre-training and downstream tasks. Such unification leads to a simplified model architecture, where only a lightweight MLM head, instead of a decoder with much more parameters, is needed on top of the multimodal encoder. Surprisingly, experimental results show that this unified framework achieves competitive performance on 14 VidL benchmarks, covering video question answering, text-to-video retrieval and video captioning. Extensive analyses further demonstrate the advantage of LAVENDER over existing VidL methods in: (i) supporting all downstream tasks with just a single set of parameter values when multi-task finetuned; (ii) few-shot generalization on various downstream tasks; and (iii) enabling zero-shot evaluation on video question answering tasks. Code is available at this https URL.

53.Self-Supervised Pretraining for Differentially Private Learning ⬇️

We demonstrate self-supervised pretraining (SSP) is a scalable solution to deep learning with differential privacy (DP) regardless of the size of available public datasets in image classification. When facing the lack of public datasets, we show the features generated by SSP on only one single image enable a private classifier to obtain much better utility than the non-learned handcrafted features under the same privacy budget. When a moderate or large size public dataset is available, the features produced by SSP greatly outperform the features trained with labels on various complex private datasets under the same private budget. We also compared multiple DP-enabled training frameworks to train a private classifier on the features generated by SSP. Finally, we report a non-trivial utility 25.3% of a private ImageNet-1K dataset when $\epsilon=3$.

54.TriHorn-Net: A Model for Accurate Depth-Based 3D Hand Pose Estimation ⬇️

3D hand pose estimation methods have made significant progress recently. However, estimation accuracy is often far from sufficient for specific real-world applications, and thus there is significant room for improvement. This paper proposes TriHorn-Net, a novel model that uses specific innovations to improve hand pose estimation accuracy on depth images. The first innovation is the decomposition of the 3D hand pose estimation into the estimation of 2D joint locations in the depth image space (UV), and the estimation of their corresponding depths aided by two complementary attention maps. This decomposition prevents depth estimation, which is a more difficult task, from interfering with the UV estimations at both the prediction and feature levels. The second innovation is PixDropout, which is, to the best of our knowledge, the first appearance-based data augmentation method for hand depth images. Experimental results demonstrate that the proposed model outperforms the state-of-the-art methods on three public benchmark datasets.

55.CRISP - Reliable Uncertainty Estimation for Medical Image Segmentation ⬇️

Accurate uncertainty estimation is a critical need for the medical imaging community. A variety of methods have been proposed, all direct extensions of classification uncertainty estimations techniques. The independent pixel-wise uncertainty estimates, often based on the probabilistic interpretation of neural networks, do not take into account anatomical prior knowledge and consequently provide sub-optimal results to many segmentation tasks. For this reason, we propose CRISP a ContRastive Image Segmentation for uncertainty Prediction method. At its core, CRISP implements a contrastive method to learn a joint latent space which encodes a distribution of valid segmentations and their corresponding images. We use this joint latent space to compare predictions to thousands of latent vectors and provide anatomically consistent uncertainty maps. Comprehensive studies performed on four medical image databases involving different modalities and organs underlines the superiority of our method compared to state-of-the-art approaches.

56.How GNNs Facilitate CNNs in Mining Geometric Information from Large-Scale Medical Images ⬇️

Gigapixel medical images provide massive data, both morphological textures and spatial information, to be mined. Due to the large data scale in histology, deep learning methods play an increasingly significant role as feature extractors. Existing solutions heavily rely on convolutional neural networks (CNNs) for global pixel-level analysis, leaving the underlying local geometric structure such as the interaction between cells in the tumor microenvironment unexplored. The topological structure in medical images, as proven to be closely related to tumor evolution, can be well characterized by graphs. To obtain a more comprehensive representation for downstream oncology tasks, we propose a fusion framework for enhancing the global image-level representation captured by CNNs with the geometry of cell-level spatial information learned by graph neural networks (GNN). The fusion layer optimizes an integration between collaborative features of global images and cell graphs. Two fusion strategies have been developed: one with MLP which is simple but turns out efficient through fine-tuning, and the other with Transformer gains a champion in fusing multiple networks. We evaluate our fusion strategies on histology datasets curated from large patient cohorts of colorectal and gastric cancers for three biomarker prediction tasks. Both two models outperform plain CNNs or GNNs, reaching a consistent AUC improvement of more than 5% on various network backbones. The experimental results yield the necessity for combining image-level morphological features with cell spatial relations in medical image analysis. Codes are available at this https URL.

57.BIO-CXRNET: A Robust Multimodal Stacking Machine Learning Technique for Mortality Risk Prediction of COVID-19 Patients using Chest X-Ray Images and Clinical Data ⬇️

Fast and accurate detection of the disease can significantly help in reducing the strain on the healthcare facility of any country to reduce the mortality during any pandemic. The goal of this work is to create a multimodal system using a novel machine learning framework that uses both Chest X-ray (CXR) images and clinical data to predict severity in COVID-19 patients. In addition, the study presents a nomogram-based scoring technique for predicting the likelihood of death in high-risk patients. This study uses 25 biomarkers and CXR images in predicting the risk in 930 COVID-19 patients admitted during the first wave of COVID-19 (March-June 2020) in Italy. The proposed multimodal stacking technique produced the precision, sensitivity, and F1-score, of 89.03%, 90.44%, and 89.03%, respectively to identify low or high-risk patients. This multimodal approach improved the accuracy by 6% in comparison to the CXR image or clinical data alone. Finally, nomogram scoring system using multivariate logistic regression -- was used to stratify the mortality risk among the high-risk patients identified in the first stage. Lactate Dehydrogenase (LDH), O2 percentage, White Blood Cells (WBC) Count, Age, and C-reactive protein (CRP) were identified as useful predictor using random forest feature selection model. Five predictors parameters and a CXR image based nomogram score was developed for quantifying the probability of death and categorizing them into two risk groups: survived (<50%), and death (>=50%), respectively. The multi-modal technique was able to predict the death probability of high-risk patients with an F1 score of 92.88 %. The area under the curves for the development and validation cohorts are 0.981 and 0.939, respectively.

58.A Deep Generative Model of Neonatal Cortical Surface Development ⬇️

The neonatal cortical surface is known to be affected by preterm birth, and the subsequent changes to cortical organisation have been associated with poorer neurodevelopmental outcomes. Deep Generative models have the potential to lead to clinically interpretable models of disease, but developing these on the cortical surface is challenging since established techniques for learning convolutional filters are inappropriate on non-flat topologies. To close this gap, we implement a surface-based CycleGAN using mixture model CNNs (MoNet) to translate sphericalised neonatal cortical surface features (curvature and T1w/T2w cortical myelin) between different stages of cortical maturity. Results show our method is able to reliably predict changes in individual patterns of cortical organisation at later stages of gestation, validated by comparison to longitudinal data; and translate appearance between preterm and term gestation (> 37 weeks gestation), validated through comparison with a trained term/preterm classifier. Simulated differences in cortical maturation are consistent with observations in the literature.

59.Body Gesture Recognition to Control a Social Robot ⬇️

In this work, we propose a gesture based language to allow humans to interact with robots using their body in a natural way. We have created a new gesture detection model using neural networks and a custom dataset of humans performing a set of body gestures to train our network. Furthermore, we compare body gesture communication with other communication channels to acknowledge the importance of adding this knowledge to robots. The presented approach is extensively validated in diverse simulations and real-life experiments with non-trained volunteers. This attains remarkable results and shows that it is a valuable framework for social robotics applications, such as human robot collaboration or human-robot interaction.

60.A Survey of Detection Methods for Die Attachment and Wire Bonding Defects in Integrated Circuit Manufacturing ⬇️

Defect detection plays a vital role in the manufacturing process of integrated circuits (ICs). Die attachment and wire bonding are two steps of the manufacturing process that determine the power and signal transmission quality and dependability in an IC. This paper presents a survey or literature review of the methods used for detecting these defects based on different sensing modalities used including optical, radiological, acoustical, and infrared thermography. A discussion of the detection methods used is provided in this survey. Both conventional and deep learning approaches for detecting die attachment and wire bonding defects are considered along with challenges and future research directions.

61.Deep Neural Network Pruning for Nuclei Instance Segmentation in Hematoxylin & Eosin-Stained Histological Images ⬇️

Recently, pruning deep neural networks (DNNs) has received a lot of attention for improving accuracy and generalization power, reducing network size, and increasing inference speed on specialized hardwares. Although pruning was mainly tested on computer vision tasks, its application in the context of medical image analysis has hardly been explored. This work investigates the impact of well-known pruning techniques, namely layer-wise and network-wide magnitude pruning, on the nuclei instance segmentation performance in histological images. Our utilized instance segmentation model consists of two main branches: (1) a semantic segmentation branch, and (2) a deep regression branch. We investigate the impact of weight pruning on the performance of both branches separately and on the final nuclei instance segmentation result. Evaluated on two publicly available datasets, our results show that layer-wise pruning delivers slightly better performance than networkwide pruning for small compression ratios (CRs) while for large CRs, network-wide pruning yields superior performance. For semantic segmentation, deep regression and final instance segmentation, 93.75 %, 95 %, and 80 % of the model weights can be pruned by layer-wise pruning with less than 2 % reduction in the performance of respective models.

62.Interpretable differential diagnosis for Alzheimer's disease and Frontotemporal dementia ⬇️

Alzheimer's disease and Frontotemporal dementia are two major types of dementia. Their accurate diagnosis and differentiation is crucial for determining specific intervention and treatment. However, differential diagnosis of these two types of dementia remains difficult at the early stage of disease due to similar patterns of clinical symptoms. Therefore, the automatic classification of multiple types of dementia has an important clinical value. So far, this challenge has not been actively explored. Recent development of deep learning in the field of medical image has demonstrated high performance for various classification tasks. In this paper, we propose to take advantage of two types of biomarkers: structure grading and structure atrophy. To this end, we propose first to train a large ensemble of 3D U-Nets to locally discriminate healthy versus dementia anatomical patterns. The result of these models is an interpretable 3D grading map capable of indicating abnormal brain regions. This map can also be exploited in various classification tasks using graph convolutional neural network. Finally, we propose to combine deep grading and atrophy-based classifications to improve dementia type discrimination. The proposed framework showed competitive performance compared to state-of-the-art methods for different tasks of disease detection and differential diagnosis.

63.Subsurface Depths Structure Maps Reconstruction with Generative Adversarial Networks ⬇️

This paper described a method for reconstruction of detailed-resolution depth structure maps, usually obtained after the 3D seismic surveys, using the data from 2D seismic depth maps. The method uses two algorithms based on the generative-adversarial neural network architecture. The first algorithm StyleGAN2-ADA accumulates in the hidden space of the neural network the semantic images of mountainous terrain forms first, and then with help of transfer learning, in the ideal case - the structure geometry of stratigraphic horizons. The second algorithm, the Pixel2Style2Pixel encoder, using the semantic level of generalization of the first algorithm, learns to reconstruct the original high-resolution images from their degraded copies (super-resolution technology). There was demonstrated a methodological approach to transferring knowledge on the structural forms of stratigraphic horizon boundaries from the well-studied areas to the underexplored ones. Using the multimodal synthesis of Pixel2Style2Pixel encoder, it is proposed to create a probabilistic depth space, where each point of the project area is represented by the density of probabilistic depth distribution of equally probable reconstructed geological forms of structural images. Assessment of the reconstruction quality was carried out for two blocks. Using this method, credible detailed depth reconstructions comparable with the quality of 3D seismic maps have been obtained from 2D seismic maps.

64.The Manifold Hypothesis for Gradient-Based Explanations ⬇️

When do gradient-based explanation algorithms provide meaningful explanations? We propose a necessary criterion: their feature attributions need to be aligned with the tangent space of the data manifold. To provide evidence for this hypothesis, we introduce a framework based on variational autoencoders that allows to estimate and generate image manifolds. Through experiments across a range of different datasets -- MNIST, EMNIST, CIFAR10, X-ray pneumonia and Diabetic Retinopathy detection -- we demonstrate that the more a feature attribution is aligned with the tangent space of the data, the more structured and explanatory it tends to be. In particular, the attributions provided by popular post-hoc methods such as Integrated Gradients, SmoothGrad and Input $\times$ Gradient tend to be more strongly aligned with the data manifold than the raw gradient. As a consequence, we suggest that explanation algorithms should actively strive to align their explanations with the data manifold. In part, this can be achieved by adversarial training, which leads to better alignment across all datasets. Some form of adjustment to the model architecture or training algorithm is necessary, since we show that generalization of neural networks alone does not imply the alignment of model gradients with the data manifold.

65.Seeking Common Ground While Reserving Differences: Multiple Anatomy Collaborative Framework for Undersampled MRI Reconstruction ⬇️

Recently, deep neural networks have greatly advanced undersampled Magnetic Resonance Image (MRI) reconstruction, wherein most studies follow the one-anatomy-one-network fashion, i.e., each expert network is trained and evaluated for a specific anatomy. Apart from inefficiency in training multiple independent models, such convention ignores the shared de-aliasing knowledge across various anatomies which can benefit each other. To explore the shared knowledge, one naive way is to combine all the data from various anatomies to train an all-round network. Unfortunately, despite the existence of the shared de-aliasing knowledge, we reveal that the exclusive knowledge across different anatomies can deteriorate specific reconstruction targets, yielding overall performance degradation. Observing this, in this study, we present a novel deep MRI reconstruction framework with both anatomy-shared and anatomy-specific parameterized learners, aiming to "seek common ground while reserving differences" across different anatomies.Particularly, the primary anatomy-shared learners are exposed to different anatomies to model flourishing shared knowledge, while the efficient anatomy-specific learners are trained with their target anatomy for exclusive knowledge. Four different implementations of anatomy-specific learners are presented and explored on the top of our framework in two MRI reconstruction networks. Comprehensive experiments on brain, knee and cardiac MRI datasets demonstrate that three of these learners are able to enhance reconstruction performance via multiple anatomy collaborative learning.

66.Robust SAR ATR on MSTAR with Deep Learning Models trained on Full Synthetic MOCEM data ⬇️

The promising potential of Deep Learning for Automatic Target Recognition (ATR) on Synthetic Aperture Radar (SAR) images vanishes when considering the complexity of collecting training datasets measurements. Simulation can overcome this issue by producing synthetic training datasets. However, because of the limited representativeness of simulation, models trained in a classical way with synthetic images have limited generalization abilities when dealing with real measurement at test time. Previous works identified a set of equally promising deep-learning algorithms to tackle this issue. However, these approaches have been evaluated in a very favorable scenario with a synthetic training dataset that overfits the ground truth of the measured test data. In this work, we study the ATR problem outside of this ideal condition, which is unlikely to occur in real operational contexts. Our contribution is threefold. (1) Using the MOCEM simulator (developed by SCALIAN DS for the French MoD/DGA), we produce a synthetic MSTAR training dataset that differs significantly from the real measurements. (2) We experimentally demonstrate the limits of the state-of-the-art. (3) We show that domain randomization techniques and adversarial training can be combined to overcome this issue. We demonstrate that this approach is more robust than the state-of-the-art, with an accuracy of 75 %, while having a limited impact on computing performance during training.

67.Differentiable Top-k Classification Learning ⬇️

The top-k classification accuracy is one of the core metrics in machine learning. Here, k is conventionally a positive integer, such as 1 or 5, leading to top-1 or top-5 training objectives. In this work, we relax this assumption and optimize the model for multiple k simultaneously instead of using a single k. Leveraging recent advances in differentiable sorting and ranking, we propose a differentiable top-k cross-entropy classification loss. This allows training the network while not only considering the top-1 prediction, but also, e.g., the top-2 and top-5 predictions. We evaluate the proposed loss function for fine-tuning on state-of-the-art architectures, as well as for training from scratch. We find that relaxing k does not only produce better top-5 accuracies, but also leads to top-1 accuracy improvements. When fine-tuning publicly available ImageNet models, we achieve a new state-of-the-art for these models.

68.Super-resolution image display using diffractive decoders ⬇️

High-resolution synthesis/projection of images over a large field-of-view (FOV) is hindered by the restricted space-bandwidth-product (SBP) of wavefront modulators. We report a deep learning-enabled diffractive display design that is based on a jointly-trained pair of an electronic encoder and a diffractive optical decoder to synthesize/project super-resolved images using low-resolution wavefront modulators. The digital encoder, composed of a trained convolutional neural network (CNN), rapidly pre-processes the high-resolution images of interest so that their spatial information is encoded into low-resolution (LR) modulation patterns, projected via a low SBP wavefront modulator. The diffractive decoder processes this LR encoded information using thin transmissive layers that are structured using deep learning to all-optically synthesize and project super-resolved images at its output FOV. Our results indicate that this diffractive image display can achieve a super-resolution factor of ~4, demonstrating a ~16-fold increase in SBP. We also experimentally validate the success of this diffractive super-resolution display using 3D-printed diffractive decoders that operate at the THz spectrum. This diffractive image decoder can be scaled to operate at visible wavelengths and inspire the design of large FOV and high-resolution displays that are compact, low-power, and computationally efficient.

69.ERNAS: An Evolutionary Neural Architecture Search for Magnetic Resonance Image Reconstructions ⬇️

Magnetic resonance imaging (MRI) is one of the noninvasive imaging modalities that can produce high-quality images. However, the scan procedure is relatively slow, which causes patient discomfort and motion artifacts in images. Accelerating MRI hardware is constrained by physical and physiological limitations. A popular alternative approach to accelerated MRI is to undersample the k-space data. While undersampling speeds up the scan procedure, it generates artifacts in the images, and advanced reconstruction algorithms are needed to produce artifact-free images. Recently deep learning has emerged as a promising MRI reconstruction method to address this problem. However, straightforward adoption of the existing deep learning neural network architectures in MRI reconstructions is not usually optimal in terms of efficiency and reconstruction quality. In this work, MRI reconstruction from undersampled data was carried out using an optimized neural network using a novel evolutionary neural architecture search algorithm. Brain and knee MRI datasets show that the proposed algorithm outperforms manually designed neural network-based MR reconstruction models.

70.On Enforcing Better Conditioned Meta-Learning for Rapid Few-Shot Adaptation ⬇️

Inspired by the concept of preconditioning, we propose a novel method to increase adaptation speed for gradient-based meta-learning methods without incurring extra parameters. We demonstrate that recasting the optimization problem to a non-linear least-squares formulation provides a principled way to actively enforce a $\textit{well-conditioned}$ parameter space for meta-learning models based on the concepts of the condition number and local curvature. Our comprehensive evaluations show that the proposed method significantly outperforms its unconstrained counterpart especially during initial adaptation steps, while achieving comparable or better overall results on several few-shot classification tasks -- creating the possibility of dynamically choosing the number of adaptation steps at inference time.

71.A Projection-Based K-space Transformer Network for Undersampled Radial MRI Reconstruction with Limited Training Subjects ⬇️

The recent development of deep learning combined with compressed sensing enables fast reconstruction of undersampled MR images and has achieved state-of-the-art performance for Cartesian k-space trajectories. However, non-Cartesian trajectories such as the radial trajectory need to be transformed onto a Cartesian grid in each iteration of the network training, slowing down the training process and posing inconvenience and delay during training. Multiple iterations of nonuniform Fourier transform in the networks offset the deep learning advantage of fast inference. Current approaches typically either work on image-to-image networks or grid the non-Cartesian trajectories before the network training to avoid the repeated gridding process. However, the image-to-image networks cannot ensure the k-space data consistency in the reconstructed images and the pre-processing of non-Cartesian k-space leads to gridding errors which cannot be compensated by the network training. Inspired by the Transformer network to handle long-range dependencies in sequence transduction tasks, we propose to rearrange the radial spokes to sequential data based on the chronological order of acquisition and use the Transformer to predict unacquired radial spokes from acquired ones. We propose novel data augmentation methods to generate a large amount of training data from a limited number of subjects. The network can be generated to different anatomical structures. Experimental results show superior performance of the proposed framework compared to state-of-the-art deep neural networks.

72.Proximal Splitting Adversarial Attacks for Semantic Segmentation ⬇️

Classification has been the focal point of research on adversarial attacks, but only a few works investigate methods suited to denser prediction tasks, such as semantic segmentation. The methods proposed in these works do not accurately solve the adversarial segmentation problem and, therefore, are overoptimistic in terms of size of the perturbations required to fool models. Here, we propose a white-box attack for these models based on a proximal splitting to produce adversarial perturbations with much smaller $\ell_1$, $\ell_2$, or $\ell_\infty$ norms. Our attack can handle large numbers of constraints within a nonconvex minimization framework via an Augmented Lagrangian approach, coupled with adaptive constraint scaling and masking strategies. We demonstrate that our attack significantly outperforms previously proposed ones, as well as classification attacks that we adapted for segmentation, providing a first comprehensive benchmark for this dense task. Our results push current limits concerning robustness evaluations in segmentation tasks.

73.Measuring Representational Harms in Image Captioning ⬇️

Previous work has largely considered the fairness of image captioning systems through the underspecified lens of "bias." In contrast, we present a set of techniques for measuring five types of representational harms, as well as the resulting measurements obtained for two of the most popular image captioning datasets using a state-of-the-art image captioning system. Our goal was not to audit this image captioning system, but rather to develop normatively grounded measurement techniques, in turn providing an opportunity to reflect on the many challenges involved. We propose multiple measurement techniques for each type of harm. We argue that by doing so, we are better able to capture the multi-faceted nature of each type of harm, in turn improving the (collective) validity of the resulting measurements. Throughout, we discuss the assumptions underlying our measurement approach and point out when they do not hold.

74.Federated Multi-organ Segmentation with Partially Labeled Data ⬇️

Federated learning is an emerging paradigm allowing large-scale decentralized learning without sharing data across different data owners, which helps address the concern of data privacy in medical image analysis. However, the requirement for label consistency across clients by the existing methods largely narrows its application scope. In practice, each clinical site may only annotate certain organs of interest with partial or no overlap with other sites. Incorporating such partially labeled data into a unified federation is an unexplored problem with clinical significance and urgency. This work tackles the challenge by using a novel federated multi-encoding U-Net (Fed-MENU) method for multi-organ segmentation. In our method, a multi-encoding U-Net (MENU-Net) is proposed to extract organ-specific features through different encoding sub-networks. Each sub-network can be seen as an expert of a specific organ and trained for that client. Moreover, to encourage the organ-specific features extracted by different sub-networks to be informative and distinctive, we regularize the training of the MENU-Net by designing an auxiliary generic decoder (AGD). Extensive experiments on four public datasets show that our Fed-MENU method can effectively obtain a federated learning model using the partially labeled datasets with superior performance to other models trained by either localized or centralized learning methods. Source code will be made publicly available at the time of paper publication.

75.Self-Supervision on Images and Text Reduces Reliance on Visual Shortcut Features ⬇️

Deep learning models trained in a fully supervised manner have been shown to rely on so-called "shortcut" features. Shortcut features are inputs that are associated with the outcome of interest in the training data, but are either no longer associated or not present in testing or deployment settings. Here we provide experiments that show recent self-supervised models trained on images and text provide more robust image representations and reduce the model's reliance on visual shortcut features on a realistic medical imaging example. Additionally, we find that these self-supervised models "forget" shortcut features more quickly than fully supervised ones when fine-tuned on labeled data. Though not a complete solution, our experiments provide compelling evidence that self-supervised models trained on images and text provide some resilience to visual shortcut features.

76.It's Time for Artistic Correspondence in Music and Video ⬇️

We present an approach for recommending a music track for a given video, and vice versa, based on both their temporal alignment and their correspondence at an artistic level. We propose a self-supervised approach that learns this correspondence directly from data, without any need of human annotations. In order to capture the high-level concepts that are required to solve the task, we propose modeling the long-term temporal context of both the video and the music signals, using Transformer networks for each modality. Experiments show that this approach strongly outperforms alternatives that do not exploit the temporal context. The combination of our contributions improve retrieval accuracy up to 10x over prior state of the art. This strong improvement allows us to introduce a wide range of analyses and applications. For instance, we can condition music retrieval based on visually defined attributes.

77.Automatic Clipping: Differentially Private Deep Learning Made Easier and Stronger ⬇️

Per-example gradient clipping is a key algorithmic step that enables practical differential private (DP) training for deep learning models. The choice of clipping norm $R$, however, is shown to be vital for achieving high accuracy under DP. We propose an easy-to-use replacement, called AutoClipping, that eliminates the need to tune $R$ for any DP optimizers, including DP-SGD, DP-Adam, DP-LAMB and many others. The automatic variants are as private and computationally efficient as existing DP optimizers, but require no DP-specific hyperparameters and thus make DP training as amenable as the standard non-private training. We give a rigorous convergence analysis of automatic DP-SGD in the non-convex setting, which shows that it enjoys an asymptotic convergence rate that matches the standard SGD. We also demonstrate on various language and vision tasks that automatic clipping outperforms or matches the state-of-the-art, and can be easily employed with minimal changes to existing codebases.

78.Loss Functions for Classification using Structured Entropy ⬇️

Cross-entropy loss is the standard metric used to train classification models in deep learning and gradient boosting. It is well-known that this loss function fails to account for similarities between the different values of the target. We propose a generalization of entropy called {\em structured entropy} which uses a random partition to incorporate the structure of the target variable in a manner which retains many theoretical properties of standard entropy. We show that a structured cross-entropy loss yields better results on several classification problems where the target variable has an a priori known structure. The approach is simple, flexible, easily computable, and does not rely on a hierarchically defined notion of structure.

79.Applications of Generative Adversarial Networks in Neuroimaging and Clinical Neuroscience ⬇️

Generative adversarial networks (GANs) are one powerful type of deep learning models that have been successfully utilized in numerous fields. They belong to a broader family called generative methods, which generate new data with a probabilistic model by learning sample distribution from real examples. In the clinical context, GANs have shown enhanced capabilities in capturing spatially complex, nonlinear, and potentially subtle disease effects compared to traditional generative methods. This review appraises the existing literature on the applications of GANs in imaging studies of various neurological conditions, including Alzheimer's disease, brain tumors, brain aging, and multiple sclerosis. We provide an intuitive explanation of various GAN methods for each application and further discuss the main challenges, open questions, and promising future directions of leveraging GANs in neuroimaging. We aim to bridge the gap between advanced deep learning methods and neurology research by highlighting how GANs can be leveraged to support clinical decision making and contribute to a better understanding of the structural and functional patterns of brain diseases.