1.Fairness Indicators for Systematic Assessments of Visual Feature Extractors ⬇️
Does everyone equally benefit from computer vision systems? Answers to this question become more and more important as computer vision systems are deployed at large scale, and can spark major concerns when they exhibit vast performance discrepancies between people from various demographic and social backgrounds. Systematic diagnosis of fairness, harms, and biases of computer vision systems is an important step towards building socially responsible systems. To initiate an effort towards standardized fairness audits, we propose three fairness indicators, which aim at quantifying harms and biases of visual systems. Our indicators use existing publicly available datasets collected for fairness evaluations, and focus on three main types of harms and bias identified in the literature, namely harmful label associations, disparity in learned representations of social and demographic traits, and biased performance on geographically diverse images from across the world.We define precise experimental protocols applicable to a wide range of computer vision models. These indicators are part of an ever-evolving suite of fairness probes and are not intended to be a substitute for a thorough analysis of the broader impact of the new computer vision technologies. Yet, we believe it is a necessary first step towards (1) facilitating the widespread adoption and mandate of the fairness assessments in computer vision research, and (2) tracking progress towards building socially responsible models. To study the practical effectiveness and broad applicability of our proposed indicators to any visual system, we apply them to off-the-shelf models built using widely adopted model training paradigms which vary in their ability to whether they can predict labels on a given image or only produce the embeddings. We also systematically study the effect of data domain and model size.
2.On Representation Learning with Feedback ⬇️
This note complements the author's recent paper "Robust representation learning with feedback for single image deraining" by providing heuristically theoretical explanations on the mechanism of representation learning with feedback, namely an essential merit of the works presented in this recent article. This note facilitates understanding of key points in the mechanism of representation learning with feedback.
3.ScoreNet: Learning Non-Uniform Attention and Augmentation for Transformer-Based Histopathological Image Classification ⬇️
Progress in digital pathology is hindered by high-resolution images and the prohibitive cost of exhaustive localized annotations. The commonly used paradigm to categorize pathology images is patch-based processing, which often incorporates multiple instance learning (MIL) to aggregate local patch-level representations yielding image-level prediction. Nonetheless, diagnostically relevant regions may only take a small fraction of the whole tissue, and MIL-based aggregation operation assumes that all patch representations are independent and thus mislays the contextual information from adjacent cell and tissue microenvironments. Consequently, the computational resources dedicated to a specific region are independent of its information contribution. This paper proposes a transformer-based architecture specifically tailored for histopathological image classification, which combines fine-grained local attention with a coarse global attention mechanism to learn meaningful representations of high-resolution images at an efficient computational cost. More importantly, based on the observation above, we propose a novel mixing-based data-augmentation strategy, namely ScoreMix, by leveraging the distribution of the semantic regions of images during the training and carefully guiding the data mixing via sampling the locations of discriminative image content. Thorough experiments and ablation studies on three challenging representative cohorts of Haematoxylin & Eosin (H&E) tumour regions-of-interest (TRoIs) datasets have validated the superiority of our approach over existing state-of-the-art methods and effectiveness of our proposed components, e.g., data augmentation in improving classification performance. We also demonstrate our method's interpretability, robustness, and cross-domain generalization capability.
4.BED: A Real-Time Object Detection System for Edge Devices ⬇️
Deploying machine learning models to edge devices has many real-world applications, especially for the scenarios that demand low latency, low power, or data privacy. However, it requires substantial research and engineering efforts due to the limited computational resources and memory of edge devices. In this demo, we present BED, an object detection system for edge devices practiced on the MAX78000 DNN accelerator. BED integrates on-device DNN inference with a camera and a screen for image acquisition and output exhibition, respectively. Experiment results indicate BED can provide accurate detection with an only 300KB tiny DNN model.
5.Texture Aware Autoencoder Pre-training And Pairwise Learning Refinement For Improved Iris Recognition ⬇️
This paper presents a texture aware end-to-end trainable iris recognition system, specifically designed for datasets like iris having limited training data. We build upon our previous stagewise learning framework with certain key optimization and architectural innovations. First, we pretrain a Stage-1 encoder network with an unsupervised autoencoder learning optimized with an additional data relation loss on top of usual reconstruction loss. The data relation loss enables learning better texture representation which is pivotal for a texture rich dataset such as iris. Robustness of Stage-1 feature representation is further enhanced with an auxiliary denoising task. Such pre-training proves beneficial for effectively training deep networks on data constrained iris datasets. Next, in Stage-2 supervised refinement, we design a pairwise learning architecture for an end-to-end trainable iris recognition system. The pairwise learning includes the task of iris matching inside the training pipeline itself and results in significant improvement in recognition performance compared to usual offline matching. We validate our model across three publicly available iris datasets and the proposed model consistently outperforms both traditional and deep learning baselines for both Within-Dataset and Cross-Dataset configurations
6.DualConv: Dual Convolutional Kernels for Lightweight Deep Neural Networks ⬇️
CNN architectures are generally heavy on memory and computational requirements which makes them infeasible for embedded systems with limited hardware resources. We propose dual convolutional kernels (DualConv) for constructing lightweight deep neural networks. DualConv combines 3$\times$3 and 1$\times$1 convolutional kernels to process the same input feature map channels simultaneously and exploits the group convolution technique to efficiently arrange convolutional filters. DualConv can be employed in any CNN model such as VGG-16 and ResNet-50 for image classification, YOLO and R-CNN for object detection, or FCN for semantic segmentation. In this paper, we extensively test DualConv for classification since these network architectures form the backbones for many other tasks. We also test DualConv for image detection on YOLO-V3. Experimental results show that, combined with our structural innovations, DualConv significantly reduces the computational cost and number of parameters of deep neural networks while surprisingly achieving slightly higher accuracy than the original models in some cases. We use DualConv to further reduce the number of parameters of the lightweight MobileNetV2 by 54% with only 0.68% drop in accuracy on CIFAR-100 dataset. When the number of parameters is not an issue, DualConv increases the accuracy of MobileNetV1 by 4.11% on the same dataset. Furthermore, DualConv significantly improves the YOLO-V3 object detection speed and improves its accuracy by 4.4% on PASCAL VOC dataset.
7.Do Lessons from Metric Learning Generalize to Image-Caption Retrieval? ⬇️
The triplet loss with semi-hard negatives has become the de facto choice for image-caption retrieval (ICR) methods that are optimized from scratch. Recent progress in metric learning has given rise to new loss functions that outperform the triplet loss on tasks such as image retrieval and representation learning. We ask whether these findings generalize to the setting of ICR by comparing three loss functions on two ICR methods. We answer this question negatively: the triplet loss with semi-hard negative mining still outperforms newly introduced loss functions from metric learning on the ICR task. To gain a better understanding of these outcomes, we introduce an analysis method to compare loss functions by counting how many samples contribute to the gradient w.r.t. the query representation during optimization. We find that loss functions that result in lower evaluation scores on the ICR task, in general, take too many (non-informative) samples into account when computing a gradient w.r.t. the query representation, which results in sub-optimal performance. The triplet loss with semi-hard negatives is shown to outperform the other loss functions, as it only takes one (hard) negative into account when computing the gradient.
8.A Survey on Image Deblurring ⬇️
With the improvement of social life quality and the real needs of daily work, images are more and more all around us. Image blurring due to camera shake, human movement, etc. has become the key to affecting image quality. How to remove image blur and restore clear image has gradually become an important research direction in the field of computer vision. After more than half a century of unremitting efforts, the majority of scientific and technological workers have made fruitful progress in image deblurring. This article reviews the work of image deblurring and specifically introduces more classic image deblurring methods, which is helpful to understand current research and look forward to future trends. This article reviews the traditional image deblurring methods and depth-represented image deblurring methods, and comprehensively classifies and introduces the corresponding technical methods. This review can provide some guidance for researchers in the field of image deblurring, and at the same time facilitate their subsequent study and research.
9.Random Walks for Adversarial Meshes ⬇️
A polygonal mesh is the most-commonly used representation of surfaces in computer graphics; thus, a variety of classification networks have been recently proposed. However, while adversarial attacks are wildly researched in 2D, almost no works on adversarial meshes exist. This paper proposes a novel, unified, and general adversarial attack, which leads to misclassification of numerous state-of-the-art mesh classification neural networks. Our attack approach is black-box, i.e. it has access only to the network's predictions, but not to the network's full architecture or gradients. The key idea is to train a network to imitate a given classification network. This is done by utilizing random walks along the mesh surface, which gather geometric information. These walks provide insight onto the regions of the mesh that are important for the correct prediction of the given classification network. These mesh regions are then modified more than other regions in order to attack the network in a manner that is barely visible to the naked eye.
10.An Automated Analysis Framework for Trajectory Datasets ⬇️
Trajectory datasets of road users have become more important in the last years for safety validation of highly automated vehicles. Several naturalistic trajectory datasets with each more than 10.000 tracks were released and others will follow. Considering this amount of data, it is necessary to be able to compare these datasets in-depth with ease to get an overview. By now, the datasets' own provided information is mainly limited to meta-data and qualitative descriptions which are mostly not consistent with other datasets. This is insufficient for users to differentiate the emerging datasets for application-specific selection. Therefore, an automated analysis framework is proposed in this work. Starting with analyzing individual tracks, fourteen elementary characteristics, so-called detection types, are derived and used as the base of this framework. To describe each traffic scenario precisely, the detections are subdivided into common metrics, clustering methods and anomaly detection. Those are combined using a modular approach. The detections are composed into new scores to describe three defined attributes of each track data quantitatively: interaction, anomaly and relevance. These three scores are calculated hierarchically for different abstract layers to provide an overview not just between datasets but also for tracks, spatial regions and individual situations. So, an objective comparison between datasets can be realized. Furthermore, it can help to get a deeper understanding of the recorded infrastructure and its effect on road user behavior. To test the validity of the framework, a study is conducted to compare the scores with human perception. Additionally, several datasets are compared.
11.Mathematical Cookbook for Snapshot Compressive Imaging ⬇️
The author intends to provide you with a beautiful, elegant, user-friendly cookbook for mathematics in Snapshot Compressive Imaging (SCI). Currently, the cookbook is composed of introduction and conventional optimization, using regularization-based optimization algorithms for SCI. The latest releases are strongly recommended! For any other questions, suggestions, or comments, feel free to email the author.
12.A precortical module for robust CNNs to light variations ⬇️
We present a simple mathematical model for the mammalian low visual pathway, taking into account its key elements: retina, lateral geniculate nucleus (LGN), primary visual cortex (V1). The analogies between the cortical level of the visual system and the structure of popular CNNs, used in image classification tasks, suggests the introduction of an additional preliminary convolutional module inspired to precortical neuronal circuits to improve robustness with respect to global light intensity and contrast variations in the input images. We validate our hypothesis on the popular databases MNIST, FashionMNIST and SVHN, obtaining significantly more robust CNNs with respect to these variations, once such extra module is added.
13.On the Complementarity of Images and Text for the Expression of Emotions in Social Media ⬇️
Authors of posts in social media communicate their emotions and what causes them with text and images. While there is work on emotion and stimulus detection for each modality separately, it is yet unknown if the modalities contain complementary emotion information in social media. We aim at filling this research gap and contribute a novel, annotated corpus of English multimodal Reddit posts. On this resource, we develop models to automatically detect the relation between image and text, an emotion stimulus category and the emotion class. We evaluate if these tasks require both modalities and find for the image-text relations, that text alone is sufficient for most categories (complementary, illustrative, opposing): the information in the text allows to predict if an image is required for emotion understanding. The emotions of anger and sadness are best predicted with a multimodal model, while text alone is sufficient for disgust, joy, and surprise. Stimuli depicted by objects, animals, food, or a person are best predicted by image-only models, while multimodal models are most effective on art, events, memes, places, or screenshots.
14.SODAR: Segmenting Objects by DynamicallyAggregating Neighboring Mask Representations ⬇️
Recent state-of-the-art one-stage instance segmentation model SOLO divides the input image into a grid and directly predicts per grid cell object masks with fully-convolutional networks, yielding comparably good performance as traditional two-stage Mask R-CNN yet enjoying much simpler architecture and higher efficiency. We observe SOLO generates similar masks for an object at nearby grid cells, and these neighboring predictions can complement each other as some may better segment certain object part, most of which are however directly discarded by non-maximum-suppression. Motivated by the observed gap, we develop a novel learning-based aggregation method that improves upon SOLO by leveraging the rich neighboring information while maintaining the architectural efficiency. The resulting model is named SODAR. Unlike the original per grid cell object masks, SODAR is implicitly supervised to learn mask representations that encode geometric structure of nearby objects and complement adjacent representations with context. The aggregation method further includes two novel designs: 1) a mask interpolation mechanism that enables the model to generate much fewer mask representations by sharing neighboring representations among nearby grid cells, and thus saves computation and memory; 2) a deformable neighbour sampling mechanism that allows the model to adaptively adjust neighbor sampling locations thus gathering mask representations with more relevant context and achieving higher performance. SODAR significantly improves the instance segmentation performance, e.g., it outperforms a SOLO model with ResNet-101 backbone by 2.2 AP on COCO \texttt{test} set, with only about 3% additional computation. We further show consistent performance gain with the SOLOv2 model.
15.Deep Learning-based Anomaly Detection on X-ray Images of Fuel Cell Electrodes ⬇️
Anomaly detection in X-ray images has been an active and lasting research area in the last decades, especially in the domain of medical X-ray images. For this work, we created a real-world labeled anomaly dataset, consisting of 16-bit X-ray image data of fuel cell electrodes coated with a platinum catalyst solution and perform anomaly detection on the dataset using a deep learning approach. The dataset contains a diverse set of anomalies with 11 identified common anomalies where the electrodes contain e.g. scratches, bubbles, smudges etc. We experiment with 16-bit image to 8-bit image conversion methods to utilize pre-trained Convolutional Neural Networks as feature extractors (transfer learning) and find that we achieve the best performance by maximizing the contrasts globally across the dataset during the 16-bit to 8-bit conversion, through histogram equalization. We group the fuel cell electrodes with anomalies into a single class called abnormal and the normal fuel cell electrodes into a class called normal, thereby abstracting the anomaly detection problem into a binary classification problem. We achieve a balanced accuracy of 85.18%. The anomaly detection is used by the company, Serenergy, for optimizing the time spend on the quality control of the fuel cell electrodes
16.A Unified Framework for Masked and Mask-Free Face Recognition via Feature Rectification ⬇️
Face recognition under ideal conditions is now considered a well-solved problem with advances in deep learning. Recognizing faces under occlusion, however, still remains a challenge. Existing techniques often fail to recognize faces with both the mouth and nose covered by a mask, which is now very common under the COVID-19 pandemic. Common approaches to tackle this problem include 1) discarding information from the masked regions during recognition and 2) restoring the masked regions before recognition. Very few works considered the consistency between features extracted from masked faces and from their mask-free counterparts. This resulted in models trained for recognizing masked faces often showing degraded performance on mask-free faces. In this paper, we propose a unified framework, named Face Feature Rectification Network (FFR-Net), for recognizing both masked and mask-free faces alike. We introduce rectification blocks to rectify features extracted by a state-of-the-art recognition model, in both spatial and channel dimensions, to minimize the distance between a masked face and its mask-free counterpart in the rectified feature space. Experiments show that our unified framework can learn a rectified feature space for recognizing both masked and mask-free faces effectively, achieving state-of-the-art results. Project code: this https URL
17.Using Social Media Images for Building Function Classification ⬇️
Urban land use on a building instance level is crucial geo-information for many applications, yet difficult to obtain. An intuitive approach to close this gap is predicting building functions from ground level imagery. Social media image platforms contain billions of images, with a large variety of motifs including but not limited to street perspectives. To cope with this issue this study proposes a filtering pipeline to yield high quality, ground level imagery from large social media image datasets. The pipeline ensures that all resulting images have full and valid geotags with a compass direction to relate image content and spatial objects from maps.
We analyze our method on a culturally diverse social media dataset from Flickr with more than 28 million images from 42 cities around the world. The obtained dataset is then evaluated in a context of 3-classes building function classification task. The three building classes that are considered in this study are: commercial, residential, and other. Fine-tuned state-of-the-art architectures yield F1-scores of up to 0.51 on the filtered images. Our analysis shows that the performance is highly limited by the quality of the labels obtained from OpenStreetMap, as the metrics increase by 0.2 if only human validated labels are considered. Therefore, we consider these labels to be weak and publish the resulting images from our pipeline together with the buildings they are showing as a weakly labeled dataset.
18.HAA4D: Few-Shot Human Atomic Action Recognition via 3D Spatio-Temporal Skeletal Alignment ⬇️
Human actions involve complex pose variations and their 2D projections can be highly ambiguous. Thus 3D spatio-temporal or 4D (i.e., 3D+T) human skeletons, which are photometric and viewpoint invariant, are an excellent alternative to 2D+T skeletons/pixels to improve action recognition accuracy. This paper proposes a new 4D dataset HAA4D which consists of more than 3,300 RGB videos in 300 human atomic action classes. HAA4D is clean, diverse, class-balanced where each class is viewpoint-balanced with the use of 4D skeletons, in which as few as one 4D skeleton per class is sufficient for training a deep recognition model. Further, the choice of atomic actions makes annotation even easier, because each video clip lasts for only a few seconds. All training and testing 3D skeletons in HAA4D are globally aligned, using a deep alignment model to the same global space, making each skeleton face the negative z-direction. Such alignment makes matching skeletons more stable by reducing intraclass variations and thus with fewer training samples per class needed for action recognition. Given the high diversity and skeletal alignment in HAA4D, we construct the first baseline few-shot 4D human atomic action recognition network without bells and whistles, which produces comparable or higher performance than relevant state-of-the-art techniques relying on embedded space encoding without explicit skeletal alignment, using the same small number of training samples of unseen classes.
19.ViNTER: Image Narrative Generation with Emotion-Arc-Aware Transformer ⬇️
Image narrative generation describes the creation of stories regarding the content of image data from a subjective viewpoint. Given the importance of the subjective feelings of writers, characters, and readers in storytelling, image narrative generation methods must consider human emotion, which is their major difference from descriptive caption generation tasks. The development of automated methods to generate story-like text associated with images may be considered to be of considerable social significance, because stories serve essential functions both as entertainment and also for many practical purposes such as education and advertising. In this study, we propose a model called ViNTER (Visual Narrative Transformer with Emotion arc Representation) to generate image narratives that focus on time series representing varying emotions as "emotion arcs," to take advantage of recent advances in multimodal Transformer-based pre-trained models. We present experimental results of both manual and automatic evaluations, which demonstrate the effectiveness of the proposed emotion-aware approach to image narrative generation.
20.Beyond Natural Motion: Exploring Discontinuity for Video Frame Interpolation ⬇️
Video interpolation is the task that synthesizes the intermediate frame given two consecutive frames. Most of the previous studies have focused on appropriate frame warping operations and refinement modules for the warped frames. These studies have been conducted on natural videos having only continuous motions. However, many practical videos contain a lot of discontinuous motions, such as chat windows, watermarks, GUI elements, or subtitles. We propose three techniques to expand the concept of transition between two consecutive frames to address these issues. First is a new architecture that can separate continuous and discontinuous motion areas. We also propose a novel data augmentation strategy called figure-text mixing (FTM) to make our model learn more general scenarios. Finally, we propose loss functions to give supervisions of the discontinuous motion areas with the data augmentation. We collected a special dataset consisting of some mobile games and chatting videos. We show that our method significantly improves the interpolation qualities of the videos on the special dataset. Moreover, our model outperforms the state-of-the-art methods for natural video datasets containing only continuous motions, such as DAVIS and UCF101.
21.Hyper-relationship Learning Network for Scene Graph Generation ⬇️
Generating informative scene graphs from images requires integrating and reasoning from various graph components, i.e., objects and relationships. However, current scene graph generation (SGG) methods, including the unbiased SGG methods, still struggle to predict informative relationships due to the lack of 1) high-level inference such as transitive inference between relationships and 2) efficient mechanisms that can incorporate all interactions of graph components. To address the issues mentioned above, we devise a hyper-relationship learning network, termed HLN, for SGG. Specifically, the proposed HLN stems from hypergraphs and two graph attention networks (GATs) are designed to infer relationships: 1) the object-relationship GAT or OR-GAT to explore interactions between objects and relationships, and 2) the hyper-relationship GAT or HR-GAT to integrate transitive inference of hyper-relationships, i.e., the sequential relationships between three objects for transitive reasoning. As a result, HLN significantly improves the performance of scene graph generation by integrating and reasoning from object interactions, relationship interactions, and transitive inference of hyper-relationships. We evaluate HLN on the most popular SGG dataset, i.e., the Visual Genome dataset, and the experimental results demonstrate its great superiority over recent state-of-the-art methods. For example, the proposed HLN improves the recall per relationship from 11.3% to 13.1%, and maintains the recall per image from 19.8% to 34.9%. We will release the source code and pretrained models on GitHub.
22.Exploring the Devil in Graph Spectral Domain for 3D Point Cloud Attacks ⬇️
3D dynamic point clouds provide a discrete representation of real-world objects or scenes in motion, which have been widely applied in immersive telepresence, autonomous driving, surveillance, \textit{etc}. However, point clouds acquired from sensors are usually perturbed by noise, which affects downstream tasks such as surface reconstruction and analysis. Although many efforts have been made for static point cloud denoising, few works address dynamic point cloud denoising. In this paper, we propose a novel gradient-based dynamic point cloud denoising method, exploiting the temporal correspondence for the estimation of gradient fields -- also a fundamental problem in dynamic point cloud processing and analysis. The gradient field is the gradient of the log-probability function of the noisy point cloud, based on which we perform gradient ascent so as to converge each point to the underlying clean surface. We estimate the gradient of each surface patch by exploiting the temporal correspondence, where the temporally corresponding patches are searched leveraging on rigid motion in classical mechanics. In particular, we treat each patch as a rigid object, which moves in the gradient field of an adjacent frame via force until reaching a balanced state, i.e., when the sum of gradients over the patch reaches 0. Since the gradient would be smaller when the point is closer to the underlying surface, the balanced patch would fit the underlying surface well, thus leading to the temporal correspondence. Finally, the position of each point in the patch is updated along the direction of the gradient averaged from corresponding patches in adjacent frames. Experimental results demonstrate that the proposed model outperforms state-of-the-art methods.
23.Review of the Fingerprint Liveness Detection (LivDet) competition series: from 2009 to 2021 ⬇️
Fingerprint authentication systems are highly vulnerable to artificial reproductions of fingerprint, called fingerprint presentation attacks. Detecting presentation attacks is not trivial because attackers refine their replication techniques from year to year. The International Fingerprint liveness Detection Competition (LivDet), an open and well-acknowledged meeting point of academies and private companies that deal with the problem of presentation attack detection, has the goal to assess the performance of fingerprint presentation attack detection (FPAD) algorithms by using standard experimental protocols and data sets. Each LivDet edition, held biannually since 2009, is characterized by a different set of challenges against which competitors must be dealt with. The continuous increase of competitors and the noticeable decrease in error rates across competitions demonstrate a growing interest in the topic. This paper reviews the LivDet editions from 2009 to 2021 and points out their evolution over the years.
24.CommerceMM: Large-Scale Commerce MultiModal Representation Learning with Omni Retrieval ⬇️
We introduce CommerceMM - a multimodal model capable of providing a diverse and granular understanding of commerce topics associated to the given piece of content (image, text, image+text), and having the capability to generalize to a wide range of tasks, including Multimodal Categorization, Image-Text Retrieval, Query-to-Product Retrieval, Image-to-Product Retrieval, etc. We follow the pre-training + fine-tuning training regime and present 5 effective pre-training tasks on image-text pairs. To embrace more common and diverse commerce data with text-to-multimodal, image-to-multimodal, and multimodal-to-multimodal mapping, we propose another 9 novel cross-modal and cross-pair retrieval tasks, called Omni-Retrieval pre-training. The pre-training is conducted in an efficient manner with only two forward/backward updates for the combined 14 tasks. Extensive experiments and analysis show the effectiveness of each task. When combining all pre-training tasks, our model achieves state-of-the-art performance on 7 commerce-related downstream tasks after fine-tuning. Additionally, we propose a novel approach of modality randomization to dynamically adjust our model under different efficiency constraints.
25.Neural Architecture Search for Dense Prediction Tasks in Computer Vision ⬇️
The success of deep learning in recent years has lead to a rising demand for neural network architecture engineering. As a consequence, neural architecture search (NAS), which aims at automatically designing neural network architectures in a data-driven manner rather than manually, has evolved as a popular field of research. With the advent of weight sharing strategies across architectures, NAS has become applicable to a much wider range of problems. In particular, there are now many publications for dense prediction tasks in computer vision that require pixel-level predictions, such as semantic segmentation or object detection. These tasks come with novel challenges, such as higher memory footprints due to high-resolution data, learning multi-scale representations, longer training times, and more complex and larger neural architectures. In this manuscript, we provide an overview of NAS for dense prediction tasks by elaborating on these novel challenges and surveying ways to address them to ease future research and application of existing methods to novel problems.
26.Few-shot semantic segmentation via mask aggregation ⬇️
Few-shot semantic segmentation aims to recognize novel classes with only very few labelled data. This challenging task requires mining of the relevant relationships between the query image and the support images. Previous works have typically regarded it as a pixel-wise classification problem. Therefore, various models have been designed to explore the correlation of pixels between the query image and the support images. However, they focus only on pixel-wise correspondence and ignore the overall correlation of objects. In this paper, we introduce a mask-based classification method for addressing this problem. The mask aggregation network (MANet), which is a simple mask classification model, is proposed to simultaneously generate a fixed number of masks and their probabilities of being targets. Then, the final segmentation result is obtained by aggregating all the masks according to their locations. Experiments on both the PASCAL-5^i and COCO-20^i datasets show that our method performs comparably to the state-of-the-art pixel-based methods. This competitive performance demonstrates the potential of mask classification as an alternative baseline method in few-shot semantic segmentation. Our source code will be made available at this https URL.
27.MeshLeTemp: Leveraging the Learnable Vertex-Vertex Relationship to Generalize Human Pose and Mesh Reconstruction for In-the-Wild Scenes ⬇️
We present MeshLeTemp, a powerful method for 3D human pose and mesh reconstruction from a single image. In terms of human body priors encoding, we propose using a learnable template human mesh instead of a constant template utilized by previous state-of-the-art methods. The proposed learnable template reflects not only vertex-vertex interactions but also the human pose and body shape, being able to adapt to diverse images. We also introduce a strategy to enrich the training data that contains both 2D and 3D annotations. We conduct extensive experiments to show the generalizability of our method and the effectiveness of our data strategy. As one of our ablation studies, we adapt MeshLeTemp to another domain which is 3D hand reconstruction.
28.Balancing Domain Experts for Long-Tailed Camera-Trap Recognition ⬇️
Label distributions in camera-trap images are highly imbalanced and long-tailed, resulting in neural networks tending to be biased towards head-classes that appear frequently. Although long-tail learning has been extremely explored to address data imbalances, few studies have been conducted to consider camera-trap characteristics, such as multi-domain and multi-frame setup. Here, we propose a unified framework and introduce two datasets for long-tailed camera-trap recognition. We first design domain experts, where each expert learns to balance imperfect decision boundaries caused by data imbalances and complement each other to generate domain-balanced decision boundaries. Also, we propose a flow consistency loss to focus on moving objects, expecting class activation maps of multi-frame matches the flow with optical flow maps for input images. Moreover, two long-tailed camera-trap datasets, WCS-LT and DMZ-LT, are introduced to validate our methods. Experimental results show the effectiveness of our framework, and proposed methods outperform previous methods on recessive domain samples.
29.Improving Human Sperm Head Morphology Classification with Unsupervised Anatomical Feature Distillation ⬇️
With rising male infertility, sperm head morphology classification becomes critical for accurate and timely clinical diagnosis. Recent deep learning (DL) morphology analysis methods achieve promising benchmark results, but leave performance and robustness on the table by relying on limited and possibly noisy class labels. To address this, we introduce a new DL training framework that leverages anatomical and image priors from human sperm microscopy crops to extract useful features without additional labeling cost. Our core idea is to distill sperm head information with reliably-generated pseudo-masks and unsupervised spatial prediction tasks. The predicted foreground masks from this distillation step are then leveraged to regularize and reduce image and label noise in the tuning stage. We evaluate our new approach on two public sperm datasets and achieve state-of-the-art performances (e.g. 65.9% SCIAN accuracy and 96.5% HuSHeM accuracy).
30.Pruning Networks with Cross-Layer Ranking & k-Reciprocal Nearest Filters ⬇️
This paper focuses on filter-level network pruning. A novel pruning method, termed CLR-RNF, is proposed. We first reveal a "long-tail" long-tail pruning problem in magnitude-based weight pruning methods, and then propose a computation-aware measurement for individual weight importance, followed by a Cross-Layer Ranking (CLR) of weights to identify and remove the bottom-ranked weights. Consequently, the per-layer sparsity makes up of the pruned network structure in our filter pruning. Then, we introduce a recommendation-based filter selection scheme where each filter recommends a group of its closest filters. To pick the preserved filters from these recommended groups, we further devise a k-Reciprocal Nearest Filter (RNF) selection scheme where the selected filters fall into the intersection of these recommended groups. Both our pruned network structure and the filter selection are non-learning processes, which thus significantly reduce the pruning complexity, and differentiate our method from existing works. We conduct image classification on CIFAR-10 and ImageNet to demonstrate the superiority of our CLR-RNF over the state-of-the-arts. For example, on CIFAR-10, CLR-RNF removes 74.1% FLOPs and 95.0% parameters from VGGNet-16 with even 0.3% accuracy improvements. On ImageNet, it removes 70.2% FLOPs and 64.8% parameters from ResNet-50 with only 1.7% top-5 accuracy drops. Our project is at this https URL.
31.GAN-generated Faces Detection: A Survey and New Perspectives ⬇️
Generative Adversarial Networks (GAN) have led to the generation of very realistic face images, which have been used in fake social media accounts and other disinformation matters that can generate profound impacts. Therefore, the corresponding GAN-face detection techniques are under active development that can examine and expose such fake faces. In this work, we aim to provide a comprehensive review of recent progress in GAN-face detection. We focus on methods that can detect face images that are generated or synthesized from GAN models. We classify the existing detection works into four categories: (1) deep learning-based, (2) physical-based, (3) physiological-based methods, and (4) evaluation and comparison against human visual performance. For each category, we summarize the key ideas and connect them with method implementations. We also discuss open problems and suggest future research directions.
32.Sim-to-Real Domain Adaptation for Lane Detection and Classification in Autonomous Driving ⬇️
While supervised detection and classification frameworks in autonomous driving require large labelled datasets to converge, Unsupervised Domain Adaptation (UDA) approaches, facilitated by synthetic data generated from photo-real simulated environments, are considered low-cost and less time-consuming solutions. In this paper, we propose UDA schemes using adversarial discriminative and generative methods for lane detection and classification applications in autonomous driving. We also present Simulanes dataset generator to create a synthetic dataset that is naturalistic utilizing CARLA's vast traffic scenarios and weather conditions. The proposed UDA frameworks take the synthesized dataset with labels as the source domain, whereas the target domain is the unlabelled real-world data. Using adversarial generative and feature discriminators, the learnt models are tuned to predict the lane location and class in the target domain. The proposed techniques are evaluated using both real-world and our synthetic datasets. The results manifest that the proposed methods have shown superiority over other baseline schemes in terms of detection and classification accuracy and consistency. The ablation study reveals that the size of the simulation dataset plays important roles in the classification performance of the proposed methods. Our UDA frameworks are available at this https URL and our dataset generator is released at this https URL
33.Rethinking Network Design and Local Geometry in Point Cloud: A Simple Residual MLP Framework ⬇️
Point cloud analysis is challenging due to irregularity and unordered data structure. To capture the 3D geometries, prior works mainly rely on exploring sophisticated local geometric extractors using convolution, graph, or attention mechanisms. These methods, however, incur unfavorable latency during inference, and the performance saturates over the past few years. In this paper, we present a novel perspective on this task. We notice that detailed local geometrical information probably is not the key to point cloud analysis -- we introduce a pure residual MLP network, called PointMLP, which integrates no sophisticated local geometrical extractors but still performs very competitively. Equipped with a proposed lightweight geometric affine module, PointMLP delivers the new state-of-the-art on multiple datasets. On the real-world ScanObjectNN dataset, our method even surpasses the prior best method by 3.3% accuracy. We emphasize that PointMLP achieves this strong performance without any sophisticated operations, hence leading to a superior inference speed. Compared to most recent CurveNet, PointMLP trains 2x faster, tests 7x faster, and is more accurate on ModelNet40 benchmark. We hope our PointMLP may help the community towards a better understanding of point cloud analysis. The code is available at this https URL.
34.Discriminability-enforcing loss to improve representation learning ⬇️
During the training process, deep neural networks implicitly learn to represent the input data samples through a hierarchy of features, where the size of the hierarchy is determined by the number of layers. In this paper, we focus on enforcing the discriminative power of the high-level representations, that are typically learned by the deeper layers (closer to the output). To this end, we introduce a new loss term inspired by the Gini impurity, which is aimed at minimizing the entropy (increasing the discriminative power) of individual high-level features with respect to the class labels. Although our Gini loss induces highly-discriminative features, it does not ensure that the distribution of the high-level features matches the distribution of the classes. As such, we introduce another loss term to minimize the Kullback-Leibler divergence between the two distributions. We conduct experiments on two image classification data sets (CIFAR-100 and Caltech 101), considering multiple neural architectures ranging from convolutional networks (ResNet-17, ResNet-18, ResNet-50) to transformers (CvT). Our empirical results show that integrating our novel loss terms into the training objective consistently outperforms the models trained with cross-entropy alone.
35.Universal Adversarial Examples in Remote Sensing: Methodology and Benchmark ⬇️
Deep neural networks have achieved great success in many important remote sensing tasks. Nevertheless, their vulnerability to adversarial examples should not be neglected. In this study, we systematically analyze the universal adversarial examples in remote sensing data for the first time, without any knowledge from the victim model. Specifically, we propose a novel black-box adversarial attack method, namely Mixup-Attack, and its simple variant Mixcut-Attack, for remote sensing data. The key idea of the proposed methods is to find common vulnerabilities among different networks by attacking the features in the shallow layer of a given surrogate model. Despite their simplicity, the proposed methods can generate transferable adversarial examples that deceive most of the state-of-the-art deep neural networks in both scene classification and semantic segmentation tasks with high success rates. We further provide the generated universal adversarial examples in the dataset named UAE-RS, which is the first dataset that provides black-box adversarial samples in the remote sensing field. We hope UAE-RS may serve as a benchmark that helps researchers to design deep neural networks with strong resistance toward adversarial attacks in the remote sensing field. Codes and the UAE-RS dataset will be available online.
36.Box Supervised Video Segmentation Proposal Network ⬇️
Video Object Segmentation (VOS) has been targeted by various fully-supervised and self-supervised approaches. While fully-supervised methods demonstrate excellent results, self-supervised ones, which do not use pixel-level ground truth, attract much attention. However, self-supervised approaches pose a significant performance gap. Box-level annotations provide a balanced compromise between labeling effort and result quality for image segmentation but have not been exploited for the video domain. In this work, we propose a box-supervised video object segmentation proposal network, which takes advantage of intrinsic video properties. Our method incorporates object motion in the following way: first, motion is computed using a bidirectional temporal difference and a novel bounding box-guided motion compensation. Second, we introduce a novel motion-aware affinity loss that encourages the network to predict positive pixel pairs if they share similar motion and color. The proposed method outperforms the state-of-the-art self-supervised benchmark by 16.4% and 6.9%
$\mathcal{J}$ &$\mathcal{F}$ score and the majority of fully supervised methods on the DAVIS and Youtube-VOS dataset without imposing network architectural specifications. We provide extensive tests and ablations on the datasets, demonstrating the robustness of our method.
37.Building Inspection Toolkit: Unified Evaluation and Strong Baselines for Damage Recognition ⬇️
In recent years, several companies and researchers have started to tackle the problem of damage recognition within the scope of automated inspection of built structures. While companies are neither willing to publish associated data nor models, researchers are facing the problem of data shortage on one hand and inconsistent dataset splitting with the absence of consistent metrics on the other hand. This leads to incomparable results. Therefore, we introduce the building inspection toolkit -- bikit -- which acts as a simple to use data hub containing relevant open-source datasets in the field of damage recognition. The datasets are enriched with evaluation splits and predefined metrics, suiting the specific task and their data distribution. For the sake of compatibility and to motivate researchers in this domain, we also provide a leaderboard and the possibility to share model weights with the community. As starting point we provide strong baselines for multi-target classification tasks utilizing extensive hyperparameter search using three transfer learning approaches for state-of-the-art algorithms. The toolkit and the leaderboard are available online.
38.A Survey of Visual Sensory Anomaly Detection ⬇️
Visual sensory anomaly detection (AD) is an essential problem in computer vision, which is gaining momentum recently thanks to the development of AI for good. Compared with semantic anomaly detection which detects anomaly at the label level (semantic shift), visual sensory AD detects the abnormal part of the sample (covariate shift). However, no thorough review has been provided to summarize this area for the computer vision community. In this survey, we are the first one to provide a comprehensive review of visual sensory AD and category into three levels according to the form of anomalies. Furthermore, we classify each kind of anomaly according to the level of supervision. Finally, we summarize the challenges and provide open directions for this community. All resources are available at this https URL.
39.ASC me to Do Anything: Multi-task Training for Embodied AI ⬇️
Embodied AI has seen steady progress across a diverse set of independent tasks. While these varied tasks have different end goals, the basic skills required to complete them successfully overlap significantly. In this paper, our goal is to leverage these shared skills to learn to perform multiple tasks jointly. We propose Atomic Skill Completion (ASC), an approach for multi-task training for Embodied AI, where a set of atomic skills shared across multiple tasks are composed together to perform the tasks. The key to the success of this approach is a pre-training scheme that decouples learning of the skills from the high-level tasks making joint training effective. We use ASC to train agents within the AI2-THOR environment to perform four interactive tasks jointly and find it to be remarkably effective. In a multi-task setting, ASC improves success rates by a factor of 2x on Seen scenes and 4x on Unseen scenes compared to no pre-training. Importantly, ASC enables us to train a multi-task agent that has a 52% higher Success Rate than training 4 independent single task agents. Finally, our hierarchical agents are more interpretable than traditional black-box architectures.
40.Lie Point Symmetry Data Augmentation for Neural PDE Solvers ⬇️
Neural networks are increasingly being used to solve partial differential equations (PDEs), replacing slower numerical solvers. However, a critical issue is that neural PDE solvers require high-quality ground truth data, which usually must come from the very solvers they are designed to replace. Thus, we are presented with a proverbial chicken-and-egg problem. In this paper, we present a method, which can partially alleviate this problem, by improving neural PDE solver sample complexity -- Lie point symmetry data augmentation (LPSDA). In the context of PDEs, it turns out that we are able to quantitatively derive an exhaustive list of data transformations, based on the Lie point symmetry group of the PDEs in question, something not possible in other application areas. We present this framework and demonstrate how it can easily be deployed to improve neural PDE solver sample complexity by an order of magnitude.
41.Improving the repeatability of deep learning models with Monte Carlo dropout ⬇️
The integration of artificial intelligence into clinical workflows requires reliable and robust models. Repeatability is a key attribute of model robustness. Repeatable models output predictions with low variation during independent tests carried out under similar conditions. During model development and evaluation, much attention is given to classification performance while model repeatability is rarely assessed, leading to the development of models that are unusable in clinical practice. In this work, we evaluate the repeatability of four model types (binary classification, multi-class classification, ordinal classification, and regression) on images that were acquired from the same patient during the same visit. We study the performance of binary, multi-class, ordinal, and regression models on four medical image classification tasks from public and private datasets: knee osteoarthritis, cervical cancer screening, breast density estimation, and retinopathy of prematurity. Repeatability is measured and compared on ResNet and DenseNet architectures. Moreover, we assess the impact of sampling Monte Carlo dropout predictions at test time on classification performance and repeatability. Leveraging Monte Carlo predictions significantly increased repeatability for all tasks on the binary, multi-class, and ordinal models leading to an average reduction of the 95% limits of agreement by 16% points and of the disagreement rate by 7% points. The classification accuracy improved in most settings along with the repeatability. Our results suggest that beyond about 20 Monte Carlo iterations, there is no further gain in repeatability. In addition to the higher test-retest agreement, Monte Carlo predictions were better calibrated which leads to output probabilities reflecting more accurately the true likelihood of being correctly classified.
42.Label fusion and training methods for reliable representation of inter-rater uncertainty ⬇️
Medical tasks are prone to inter-rater variability due to multiple factors such as image quality, professional experience and training, or guideline clarity. Training deep learning networks with annotations from multiple raters is a common practice that mitigates the model's bias towards a single expert. Reliable models generating calibrated outputs and reflecting the inter-rater disagreement are key to the integration of artificial intelligence in clinical practice. Various methods exist to take into account different expert labels. We focus on comparing three label fusion methods: STAPLE, average of the rater's segmentation, and random sampling each rater's segmentation during training. Each label fusion method is studied using the conventional training framework or the recently published SoftSeg framework that limits information loss by treating the segmentation task as a regression. Our results, across 10 data splittings on two public datasets, indicate that SoftSeg models, regardless of the ground truth fusion method, had better calibration and preservation of the inter-rater rater variability compared with their conventional counterparts without impacting the segmentation performance. Conventional models, i.e., trained with a Dice loss, with binary inputs, and sigmoid/softmax final activate, were overconfident and underestimated the uncertainty associated with inter-rater variability. Conversely, fusing labels by averaging with the SoftSeg framework led to underconfident outputs and overestimation of the rater disagreement. In terms of segmentation performance, the best label fusion method was different for the two datasets studied, indicating this parameter might be task-dependent. However, SoftSeg had segmentation performance systematically superior or equal to the conventionally trained models and had the best calibration and preservation of the inter-rater variability.
43.Post-Training Quantization for Cross-Platform Learned Image Compression ⬇️
It has been witnessed that learned image compression has outperformed conventional image coding techniques and tends to be practical in industrial applications. One of the most critical issues that need to be considered is the non-deterministic calculation, which makes the probability prediction cross-platform inconsistent and frustrates successful decoding. We propose to solve this problem by introducing well-developed post-training quantization and making the model inference integer-arithmetic-only, which is much simpler than presently existing training and fine-tuning based approaches yet still keeps the superior rate-distortion performance of learned image compression. Based on that, we further improve the discretization of the entropy parameters and extend the deterministic inference to fit Gaussian mixture models. With our proposed methods, the current state-of-the-art image compression models can infer in a cross-platform consistent manner, which makes the further development and practice of learned image compression more promising.
44.Deep Constrained Least Squares for Blind Image Super-Resolution ⬇️
In this paper, we tackle the problem of blind image super-resolution(SR) with a reformulated degradation model and two novel modules. Following the common practices of blind SR, our method proposes to improve both the kernel estimation as well as the kernel based high resolution image restoration. To be more specific, we first reformulate the degradation model such that the deblurring kernel estimation can be transferred into the low resolution space. On top of this, we introduce a dynamic deep linear filter module. Instead of learning a fixed kernel for all images, it can adaptively generate deblurring kernel weights conditional on the input and yields more robust kernel estimation. Subsequently, a deep constrained least square filtering module is applied to generate clean features based on the reformulation and estimated kernel. The deblurred feature and the low input image feature are then fed into a dual-path structured SR network and restore the final high resolution result. To evaluate our method, we further conduct evaluations on several benchmarks, including Gaussian8 and DIV2KRK. Our experiments demonstrate that the proposed method achieves better accuracy and visual improvements against state-of-the-art methods.
45.A Real-time System for Detecting Landslide Reports on Social Media using Artificial Intelligence ⬇️
This paper presents an online system that leverages social media data in real time to identify landslide-related information automatically using state-of-the-art artificial intelligence techniques. The designed system can (i) reduce the information overload by eliminating duplicate and irrelevant content, (ii) identify landslide images, (iii) infer geolocation of the images, and (iv) categorize the user type (organization or person) of the account sharing the information. The system was deployed in February 2020 online at this https URL to monitor live Twitter data stream and has been running continuously since then to provide time-critical information to partners such as British Geological Survey and European Mediterranean Seismological Centre. We trust this system can both contribute to harvesting of global landslide data for further research and support global landslide maps to facilitate emergency response and decision making.
46.Federated Contrastive Learning for Dermatological Disease Diagnosis via On-device Learning ⬇️
Deep learning models have been deployed in an increasing number of edge and mobile devices to provide healthcare. These models rely on training with a tremendous amount of labeled data to achieve high accuracy. However, for medical applications such as dermatological disease diagnosis, the private data collected by mobile dermatology assistants exist on distributed mobile devices of patients, and each device only has a limited amount of data. Directly learning from limited data greatly deteriorates the performance of learned models. Federated learning (FL) can train models by using data distributed on devices while keeping the data local for privacy. Existing works on FL assume all the data have ground-truth labels. However, medical data often comes without any accompanying labels since labeling requires expertise and results in prohibitively high labor costs. The recently developed self-supervised learning approach, contrastive learning (CL), can leverage the unlabeled data to pre-train a model, after which the model is fine-tuned on limited labeled data for dermatological disease diagnosis. However, simply combining CL with FL as federated contrastive learning (FCL) will result in ineffective learning since CL requires diverse data for learning but each device only has limited data. In this work, we propose an on-device FCL framework for dermatological disease diagnosis with limited labels. Features are shared in the FCL pre-training process to provide diverse and accurate contrastive information. After that, the pre-trained model is fine-tuned with local labeled data independently on each device or collaboratively with supervised federated learning on all devices. Experiments on dermatological disease datasets show that the proposed framework effectively improves the recall and precision of dermatological disease diagnosis compared with state-of-the-art methods.
47.DeepSensor: Deep Learning Testing Framework Based on Neuron Sensitivity ⬇️
Despite impressive capabilities and outstanding performance, deep neural network(DNN) has captured increasing public concern for its security problem, due to frequent occurrence of erroneous behaviors. Therefore, it is necessary to conduct systematically testing before its deployment to real-world applications. Existing testing methods have provided fine-grained criteria based on neuron coverage and reached high exploratory degree of testing. But there is still a gap between the neuron coverage and model's robustness evaluation. To bridge the gap, we observed that neurons which change the activation value dramatically due to minor perturbation are prone to trigger incorrect corner cases. Motivated by it, we propose neuron sensitivity and develop a novel white-box testing framework for DNN, donated as DeepSensor. The number of sensitive neurons is maximized by particle swarm optimization, thus diverse corner cases could be triggered and neuron coverage be further improved when compared with baselines. Besides, considerable robustness enhancement can be reached when adopting testing examples based on neuron sensitivity for retraining. Extensive experiments implemented on scalable datasets and models can well demonstrate the testing effectiveness and robustness improvement of DeepSensor.
48.Explainable COVID-19 Infections Identification and Delineation Using Calibrated Pseudo Labels ⬇️
The upheaval brought by the arrival of the COVID-19 pandemic has continued to bring fresh challenges over the past two years. During this COVID-19 pandemic, there has been a need for rapid identification of infected patients and specific delineation of infection areas in computed tomography (CT) images. Although deep supervised learning methods have been established quickly, the scarcity of both image-level and pixellevel labels as well as the lack of explainable transparency still hinder the applicability of AI. Can we identify infected patients and delineate the infections with extreme minimal supervision? Semi-supervised learning (SSL) has demonstrated promising performance under limited labelled data and sufficient unlabelled data. Inspired by SSL, we propose a model-agnostic calibrated pseudo-labelling strategy and apply it under a consistency regularization framework to generate explainable identification and delineation results. We demonstrate the effectiveness of our model with the combination of limited labelled data and sufficient unlabelled data or weakly-labelled data. Extensive experiments have shown that our model can efficiently utilize limited labelled data and provide explainable classification and segmentation results for decision-making in clinical routine.
49.Multimodal Driver Referencing: A Comparison of Pointing to Objects Inside and Outside the Vehicle ⬇️
Advanced in-cabin sensing technologies, especially vision based approaches, have tremendously progressed user interaction inside the vehicle, paving the way for new applications of natural user interaction. Just as humans use multiple modes to communicate with each other, we follow an approach which is characterized by simultaneously using multiple modalities to achieve natural human-machine interaction for a specific task: pointing to or glancing towards objects inside as well as outside the vehicle for deictic references. By tracking the movements of eye-gaze, head and finger, we design a multimodal fusion architecture using a deep neural network to precisely identify the driver's referencing intent. Additionally, we use a speech command as a trigger to separate each referencing event. We observe differences in driver behavior in the two pointing use cases (i.e. for inside and outside objects), especially when analyzing the preciseness of the three modalities eye, head, and finger. We conclude that there is no single modality that is solely optimal for all cases as each modality reveals certain limitations. Fusion of multiple modalities exploits the relevant characteristics of each modality, hence overcoming the case dependent limitations of each individual modality. Ultimately, we propose a method to identity whether the driver's referenced object lies inside or outside the vehicle, based on the predicted pointing direction.
50.A Survey of Neural Trojan Attacks and Defenses in Deep Learning ⬇️
Artificial Intelligence (AI) relies heavily on deep learning - a technology that is becoming increasingly popular in real-life applications of AI, even in the safety-critical and high-risk domains. However, it is recently discovered that deep learning can be manipulated by embedding Trojans inside it. Unfortunately, pragmatic solutions to circumvent the computational requirements of deep learning, e.g. outsourcing model training or data annotation to third parties, further add to model susceptibility to the Trojan attacks. Due to the key importance of the topic in deep learning, recent literature has seen many contributions in this direction. We conduct a comprehensive review of the techniques that devise Trojan attacks for deep learning and explore their defenses. Our informative survey systematically organizes the recent literature and discusses the key concepts of the methods while assuming minimal knowledge of the domain on the readers part. It provides a comprehensible gateway to the broader community to understand the recent developments in Neural Trojans.
51.To what extent can Plug-and-Play methods outperform neural networks alone in low-dose CT reconstruction ⬇️
The Plug-and-Play (PnP) framework was recently introduced for low-dose CT reconstruction to leverage the interpretability and the flexibility of model-based methods to incorporate various plugins, such as trained deep learning (DL) neural networks. However, the benefits of PnP vs. state-of-the-art DL methods have not been clearly demonstrated. In this work, we proposed an improved PnP framework to address the previous limitations and develop clinical-relevant segmentation metrics for quantitative result assessment. Compared with the DL alone methods, our proposed PnP framework was slightly inferior in MSE and PSNR. However, the power spectrum of the resulting images better matched that of full-dose images than that of DL denoised images. The resulting images supported higher accuracy in airway segmentation than DL denoised images for all the ten patients in the test set, more substantially on the airways with a cross-section smaller than 0.61cm$^2$, and outperformed the DL denoised images for 45 out of 50 lung lobes in lobar segmentation. Our PnP method proved to be significantly better at preserving the image texture, which translated to task-specific benefits in automated structure segmentation and detection.
52.Debiased Pseudo Labeling in Self-Training ⬇️
Deep neural networks achieve remarkable performances on a wide range of tasks with the aid of large-scale labeled datasets. However, large-scale annotations are time-consuming and labor-exhaustive to obtain on realistic tasks. To mitigate the requirement for labeled data, self-training is widely used in both academia and industry by pseudo labeling on readily-available unlabeled data. Despite its popularity, pseudo labeling is well-believed to be unreliable and often leads to training instability. Our experimental studies further reveal that the performance of self-training is biased due to data sampling, pre-trained models, and training strategies, especially the inappropriate utilization of pseudo labels. To this end, we propose Debiased, in which the generation and utilization of pseudo labels are decoupled by two independent heads. To further improve the quality of pseudo labels, we introduce a worst-case estimation of pseudo labeling and seamlessly optimize the representations to avoid the worst-case. Extensive experiments justify that the proposed Debiased not only yields an average improvement of
$14.4$ % against state-of-the-art algorithms on$11$ tasks (covering generic object recognition, fine-grained object recognition, texture classification, and scene classification) but also helps stabilize training and balance performance across classes.
53.Compositional Scene Representation Learning via Reconstruction: A Survey ⬇️
Visual scene representation learning is an important research problem in the field of computer vision. The performance on vision tasks could be improved if more suitable representations are learned for visual scenes. Complex visual scenes are the composition of relatively simple visual concepts, and have the property of combinatorial explosion. Compared with directly representing the entire visual scene, extracting compositional scene representations can better cope with the diverse combination of background and objects. Because compositional scene representations abstract the concept of objects, performing visual scene analysis and understanding based on these representations could be easier and more interpretable. Moreover, learning compositional scene representations via reconstruction can greatly reduce the need for training data annotations. Therefore, compositional scene representation learning via reconstruction has important research significance. In this survey, we first discuss representative methods that either learn from a single viewpoint or multiple viewpoints without object-level supervision, then the applications of compositional scene representations, and finally the future directions on this topic.
54.Multi-task UNet: Jointly Boosting Saliency Prediction and Disease Classification on Chest X-ray Images ⬇️
Human visual attention has recently shown its distinct capability in boosting machine learning models. However, studies that aim to facilitate medical tasks with human visual attention are still scarce. To support the use of visual attention, this paper describes a novel deep learning model for visual saliency prediction on chest X-ray (CXR) images. To cope with data deficiency, we exploit the multi-task learning method and tackles disease classification on CXR simultaneously. For a more robust training process, we propose a further optimized multi-task learning scheme to better handle model overfitting. Experiments show our proposed deep learning model with our new learning scheme can outperform existing methods dedicated either for saliency prediction or image classification. The code used in this paper is available at this https URL.
55.Gaze-Guided Class Activation Mapping: Leveraging Human Attention for Network Attention in Chest X-rays Classification ⬇️
The increased availability and accuracy of eye-gaze tracking technology has sparked attention-related research in psychology, neuroscience, and, more recently, computer vision and artificial intelligence. The attention mechanism in artificial neural networks is known to improve learning tasks. However, no previous research has combined the network attention and human attention. This paper describes a gaze-guided class activation mapping (GG-CAM) method to directly regulate the formation of network attention based on expert radiologists' visual attention for the chest X-ray pathology classification problem, which remains challenging due to the complex and often nuanced differences among images. GG-CAM is a lightweight (
$3$ additional trainable parameters for regulating the learning process) and generic extension that can be easily applied to most classification convolutional neural networks (CNN). GG-CAM-modified CNNs do not require human attention as an input when fully trained. Comparative experiments suggest that two standard CNNs with the GG-CAM extension achieve significantly greater classification performance. The median area under the curve (AUC) metrics for ResNet50 increases from$0.721$ to$0.776$ . For EfficientNetv2 (s), the median AUC increases from$0.723$ to$0.801$ . The GG-CAM also brings better interpretability of the network that facilitates the weakly-supervised pathology localization and analysis.
56.One Step at a Time: Long-Horizon Vision-and-Language Navigation with Milestones ⬇️
We study the problem of developing autonomous agents that can follow human instructions to infer and perform a sequence of actions to complete the underlying task. Significant progress has been made in recent years, especially for tasks with short horizons. However, when it comes to long-horizon tasks with extended sequences of actions, an agent can easily ignore some instructions or get stuck in the middle of the long instructions and eventually fail the task. To address this challenge, we propose a model-agnostic milestone-based task tracker (M-TRACK) to guide the agent and monitor its progress. Specifically, we propose a milestone builder that tags the instructions with navigation and interaction milestones which the agent needs to complete step by step, and a milestone checker that systemically checks the agent's progress in its current milestone and determines when to proceed to the next. On the challenging ALFRED dataset, our M-TRACK leads to a notable 45% and 70% relative improvement in unseen success rate over two competitive base models.
57.Handcrafted Histological Transformer (H2T): Unsupervised Representation of Whole Slide Images ⬇️
Diagnostic, prognostic and therapeutic decision-making of cancer in pathology clinics can now be carried out based on analysis of multi-gigapixel tissue images, also known as whole-slide images (WSIs). Recently, deep convolutional neural networks (CNNs) have been proposed to derive unsupervised WSI representations; these are attractive as they rely less on expert annotation which is cumbersome. However, a major trade-off is that higher predictive power generally comes at the cost of interpretability, posing a challenge to their clinical use where transparency in decision-making is generally expected. To address this challenge, we present a handcrafted framework based on deep CNN for constructing holistic WSI-level representations. Building on recent findings about the internal working of the Transformer in the domain of natural language processing, we break down its processes and handcraft them into a more transparent framework that we term as the Handcrafted Histological Transformer or H2T. Based on our experiments involving various datasets consisting of a total of 5,306 WSIs, the results demonstrate that H2T based holistic WSI-level representations offer competitive performance compared to recent state-of-the-art methods and can be readily utilized for various downstream analysis tasks. Finally, our results demonstrate that the H2T framework can be up to 14 times faster than the Transformer models.
58.A Survey of Cross-Modality Brain Image Synthesis ⬇️
The existence of completely aligned and paired multi-modal neuroimaging data has proved its effectiveness in diagnosis of brain diseases. However, collecting the full set of well-aligned and paired data is impractical or even luxurious, since the practical difficulties may include high cost, long time acquisition, image corruption, and privacy issues. A realistic solution is to explore either an unsupervised learning or a semi-supervised learning to synthesize the absent neuroimaging data. In this paper, we tend to approach multi-modality brain image synthesis task from different perspectives, which include the level of supervision, the range of modality synthesis, and the synthesis-based downstream tasks. Particularly, we provide in-depth analysis on how cross-modality brain image synthesis can improve the performance of different downstream tasks. Finally, we evaluate the challenges and provide several open directions for this community. All resources are available at this https URL
59.DermX: an end-to-end framework for explainable automated dermatological diagnosis ⬇️
Dermatological diagnosis automation is essential in addressing the high prevalence of skin diseases and critical shortage of dermatologists. Despite approaching expert-level diagnosis performance, convolutional neural network (ConvNet) adoption in clinical practice is impeded by their limited explainability, and by subjective, expensive explainability validations. We introduce DermX and DermX+, an end-to-end framework for explainable automated dermatological diagnosis. DermX is a clinically-inspired explainable dermatological diagnosis ConvNet, trained using DermXDB, a 554 images dataset annotated by eight dermatologists with diagnoses and supporting explanations. DermX+ extends DermX with guided attention training for explanation attention maps. Both methods achieve near-expert diagnosis performance, with DermX, DermX+, and dermatologist F1 scores of 0.79, 0.79, and 0.87, respectively. We assess the explanation plausibility in terms of identification and localization, by comparing model-selected with dermatologist-selected explanations, and gradient-weighted class-activation maps with dermatologist explanation maps. Both DermX and DermX+ obtain an identification F1 score of 0.78. The localization F1 score is 0.39 for DermX and 0.35 for DermX+. Explanation faithfulness is assessed through contrasting samples, DermX obtaining 0.53 faithfulness and DermX+ 0.25. These results show that explainability does not necessarily come at the expense of predictive power, as our high-performance models provide both plausible and faithful explanations for their diagnoses.