1.Borrowing from yourself: Faster future video segmentation with partial channel update ⬇️
Semantic segmentation is a well-addressed topic in the computer vision literature, but the design of fast and accurate video processing networks remains challenging. In addition, to run on embedded hardware, computer vision models often have to make compromises on accuracy to run at the required speed, so that a latency/accuracy trade-off is usually at the heart of these real-time systems' design. For the specific case of videos, models have the additional possibility to make use of computations made for previous frames to mitigate the accuracy loss while being real-time.
In this work, we propose to tackle the task of fast future video segmentation prediction through the use of convolutional layers with time-dependent channel masking. This technique only updates a chosen subset of the feature maps at each time-step, bringing simultaneously less computation and latency, and allowing the network to leverage previously computed features. We apply this technique to several fast architectures and experimentally confirm its benefits for the future prediction subtask.
2.Patch-NetVLAD+: Learned patch descriptor and weighted matching strategy for place recognition ⬇️
Visual Place Recognition (VPR) in areas with similar scenes such as urban or indoor scenarios is a major challenge. Existing VPR methods using global descriptors have difficulty capturing local specific regions (LSR) in the scene and are therefore prone to localization confusion in such scenarios. As a result, finding the LSR that are critical for location recognition becomes key. To address this challenge, we introduced Patch-NetVLAD+, which was inspired by patch-based VPR researches. Our method proposed a fine-tuning strategy with triplet loss to make NetVLAD suitable for extracting patch-level descriptors. Moreover, unlike existing methods that treat all patches in an image equally, our method extracts patches of LSR, which present less frequently throughout the dataset, and makes them play an important role in VPR by assigning proper weights to them. Experiments on Pittsburgh30k and Tokyo247 datasets show that our approach achieved up to 6.35% performance improvement than existing patch-based methods.
3.Deep soccer captioning with transformer: dataset, semantics-related losses, and multi-level evaluation ⬇️
This work aims at generating captions for soccer videos using deep learning. In this context, this paper introduces a dataset, model, and triple-level evaluation. The dataset consists of 22k caption-clip pairs and three visual features (images, optical flow, inpainting) for ~500 hours of \emph{SoccerNet} videos. The model is divided into three parts: a transformer learns language, ConvNets learn vision, and a fusion of linguistic and visual features generates captions. The paper suggests evaluating generated captions at three levels: syntax (the commonly used evaluation metrics such as BLEU-score and CIDEr), meaning (the quality of descriptions for a domain expert), and corpus (the diversity of generated captions). The paper shows that the diversity of generated captions has improved (from 0.07 reaching 0.18) with semantics-related losses that prioritize selected words. Semantics-related losses and the utilization of more visual features (optical flow, inpainting) improved the normalized captioning score by 28%. The web page of this work: this https URL}{this https URL
4.SuperCon: Supervised Contrastive Learning for Imbalanced Skin Lesion Classification ⬇️
Convolutional neural networks (CNNs) have achieved great success in skin lesion classification. A balanced dataset is required to train a good model. However, due to the appearance of different skin lesions in practice, severe or even deadliest skin lesion types (e.g., melanoma) naturally have quite small amount represented in a dataset. In that, classification performance degradation occurs widely, it is significantly important to have CNNs that work well on class imbalanced skin lesion image dataset. In this paper, we propose SuperCon, a two-stage training strategy to overcome the class imbalance problem on skin lesion classification. It contains two stages: (i) representation training that tries to learn a feature representation that closely aligned among intra-classes and distantly apart from inter-classes, and (ii) classifier fine-tuning that aims to learn a classifier that correctly predict the label based on the learnt representations. In the experimental evaluation, extensive comparisons have been made among our approach and other existing approaches on skin lesion benchmark datasets. The results show that our two-stage training strategy effectively addresses the class imbalance classification problem, and significantly improves existing works in terms of F1-score and AUC score, resulting in state-of-the-art performance.
5.Tiny Object Tracking: A Large-scale Dataset and A Baseline ⬇️
Tiny objects, frequently appearing in practical applications, have weak appearance and features, and receive increasing interests in meany vision tasks, such as object detection and segmentation. To promote the research and development of tiny object tracking, we create a large-scale video dataset, which contains 434 sequences with a total of more than 217K frames. Each frame is carefully annotated with a high-quality bounding box. In data creation, we take 12 challenge attributes into account to cover a broad range of viewpoints and scene complexities, and annotate these attributes for facilitating the attribute-based performance analysis. To provide a strong baseline in tiny object tracking, we propose a novel Multilevel Knowledge Distillation Network (MKDNet), which pursues three-level knowledge distillations in a unified framework to effectively enhance the feature representation, discrimination and localization abilities in tracking tiny objects. Extensive experiments are performed on the proposed dataset, and the results prove the superiority and effectiveness of MKDNet compared with state-of-the-art methods. The dataset, the algorithm code, and the evaluation code are available at this https URL.
6.Video-driven Neural Physically-based Facial Asset for Production ⬇️
Production-level workflows for producing convincing 3D dynamic human faces have long relied on a disarray of labor-intensive tools for geometry and texture generation, motion capture and rigging, and expression synthesis. Recent neural approaches automate individual components but the corresponding latent representations cannot provide artists explicit controls as in conventional tools. In this paper, we present a new learning-based, video-driven approach for generating dynamic facial geometries with high-quality physically-based assets. Two key components are well-structured latent spaces due to dense temporal samplings from videos and explicit facial expression controls to regulate the latent spaces. For data collection, we construct a hybrid multiview-photometric capture stage, coupling with an ultra-fast video camera to obtain raw 3D facial assets. We then model the facial expression, geometry and physically-based textures using separate VAEs with a global MLP-based expression mapping across the latent spaces, to preserve characteristics across respective attributes while maintaining explicit controls over geometry and texture. We also introduce to model the delta information as wrinkle maps for physically-base textures, achieving high-quality rendering of dynamic textures. We demonstrate our approach in high-fidelity performer-specific facial capture and cross-identity facial motion retargeting. In addition, our neural asset along with fast adaptation schemes can also be deployed to handle in-the-wild videos. Besides, we motivate the utility of our explicit facial disentangle strategy by providing promising physically-based editing results like geometry and material editing or winkle transfer with high realism. Comprehensive experiments show that our technique provides higher accuracy and visual fidelity than previous video-driven facial reconstruction and animation methods.
7.Towards Weakly-Supervised Text Spotting using a Multi-Task Transformer ⬇️
Text spotting end-to-end methods have recently gained attention in the literature due to the benefits of jointly optimizing the text detection and recognition components. Existing methods usually have a distinct separation between the detection and recognition branches, requiring exact annotations for the two tasks. We introduce TextTranSpotter (TTS), a transformer-based approach for text spotting and the first text spotting framework which may be trained with both fully- and weakly-supervised settings. By learning a single latent representation per word detection, and using a novel loss function based on the Hungarian loss, our method alleviates the need for expensive localization annotations. Trained with only text transcription annotations on real data, our weakly-supervised method achieves competitive performance with previous state-of-the-art fully-supervised methods. When trained in a fully-supervised manner, TextTranSpotter shows state-of-the-art results on multiple benchmarks.
8.Multi-Modal Fusion for Sensorimotor Coordination in Steering Angle Prediction ⬇️
Imitation learning is employed to learn sensorimotor coordination for steering angle prediction in an end-to-end fashion requires expert demonstrations. These expert demonstrations are paired with environmental perception and vehicle control data. The conventional frame-based RGB camera is the most common exteroceptive sensor modality used to acquire the environmental perception data. The frame-based RGB camera has produced promising results when used as a single modality in learning end-to-end lateral control. However, the conventional frame-based RGB camera has limited operability in illumination variation conditions and is affected by the motion blur. The event camera provides complementary information to the frame-based RGB camera. This work explores the fusion of frame-based RGB and event data for learning end-to-end lateral control by predicting steering angle. In addition, how the representation from event data fuse with frame-based RGB data helps to predict the lateral control robustly for the autonomous vehicle. To this end, we propose DRFuser, a novel convolutional encoder-decoder architecture for learning end-to-end lateral control. The encoder module is branched between the frame-based RGB data and event data along with the self-attention layers. Moreover, this study has also contributed to our own collected dataset comprised of event, frame-based RGB, and vehicle control data. The efficacy of the proposed method is experimentally evaluated on our collected dataset, Davis Driving dataset (DDD), and Carla Eventscape dataset. The experimental results illustrate that the proposed method DRFuser outperforms the state-of-the-art in terms of root-mean-square error (RMSE) and mean absolute error (MAE) used as evaluation metrics.
9.Exemplar-free Online Continual Learning ⬇️
Targeted for real world scenarios, online continual learning aims to learn new tasks from sequentially available data under the condition that each data is observed only once by the learner. Though recent works have made remarkable achievements by storing part of learned task data as exemplars for knowledge replay, the performance is greatly relied on the size of stored exemplars while the storage consumption is a significant constraint in continual learning. In addition, storing exemplars may not always be feasible for certain applications due to privacy concerns. In this work, we propose a novel exemplar-free method by leveraging nearest-class-mean (NCM) classifier where the class mean is estimated during training phase on all data seen so far through online mean update criteria. We focus on image classification task and conduct extensive experiments on benchmark datasets including CIFAR-100 and Food-1k. The results demonstrate that our method without using any exemplar outperforms state-of-the-art exemplar-based approaches with large margins under standard protocol (20 exemplars per class) and is able to achieve competitive performance even with larger exemplar size (100 exemplars per class).
10.Bench-Marking And Improving Arabic Automatic Image Captioning Through The Use Of Multi-Task Learning Paradigm ⬇️
The continuous increase in the use of social media and the visual content on the internet have accelerated the research in computer vision field in general and the image captioning task in specific. The process of generating a caption that best describes an image is a useful task for various applications such as it can be used in image indexing and as a hearing aid for the visually impaired. In recent years, the image captioning task has witnessed remarkable advances regarding both datasets and architectures, and as a result, the captioning quality has reached an astounding performance. However, the majority of these advances especially in datasets are targeted for English, which left other languages such as Arabic lagging behind. Although Arabic language, being spoken by more than 450 million people and being the most growing language on the internet, lacks the fundamental pillars it needs to advance its image captioning research, such as benchmarks or unified datasets. This works is an attempt to expedite the synergy in this task by providing unified datasets and benchmarks, while also exploring methods and techniques that could enhance the performance of Arabic image captioning. The use of multi-task learning is explored, alongside exploring various word representations and different features. The results showed that the use of multi-task learning and pre-trained word embeddings noticeably enhanced the quality of image captioning, however the presented results shows that Arabic captioning still lags behind when compared to the English language. The used dataset and code are available at this link.
11.WAD-CMSN: Wasserstein Distance based Cross-Modal Semantic Network for Zero-Shot Sketch-Based Image Retrieval ⬇️
Zero-shot sketch-based image retrieval (ZSSBIR), as a popular studied branch of computer vision, attracts wide attention recently. Unlike sketch-based image retrieval (SBIR), the main aim of ZSSBIR is to retrieve natural images given free hand-drawn sketches that may not appear during training. Previous approaches used semantic aligned sketch-image pairs or utilized memory expensive fusion layer for projecting the visual information to a low dimensional subspace, which ignores the significant heterogeneous cross-domain discrepancy between highly abstract sketch and relevant image. This may yield poor performance in the training phase. To tackle this issue and overcome this drawback, we propose a Wasserstein distance based cross-modal semantic network (WAD-CMSN) for ZSSBIR. Specifically, it first projects the visual information of each branch (sketch, image) to a common low dimensional semantic subspace via Wasserstein distance in an adversarial training manner. Furthermore, identity matching loss is employed to select useful features, which can not only capture complete semantic knowledge, but also alleviate the over-fitting phenomenon caused by the WAD-CMSN model. Experimental results on the challenging Sketchy (Extended) and TU-Berlin (Extended) datasets indicate the effectiveness of the proposed WAD-CMSN model over several competitors.
12.ACORT: A Compact Object Relation Transformer for Parameter Efficient Image Captioning ⬇️
Recent research that applies Transformer-based architectures to image captioning has resulted in state-of-the-art image captioning performance, capitalising on the success of Transformers on natural language tasks. Unfortunately, though these models work well, one major flaw is their large model sizes. To this end, we present three parameter reduction methods for image captioning Transformers: Radix Encoding, cross-layer parameter sharing, and attention parameter sharing. By combining these methods, our proposed ACORT models have 3.7x to 21.6x fewer parameters than the baseline model without compromising test performance. Results on the MS-COCO dataset demonstrate that our ACORT models are competitive against baselines and SOTA approaches, with CIDEr score >=126. Finally, we present qualitative results and ablation studies to demonstrate the efficacy of the proposed changes further. Code and pre-trained models are publicly available at this https URL.
13.Incremental Learning of Structured Memory via Closed-Loop Transcription ⬇️
This work proposes a minimal computational model for learning a structured memory of multiple object classes in an incremental setting. Our approach is based on establishing a closed-loop transcription between multiple classes and their corresponding subspaces, known as a linear discriminative representation, in a low-dimensional feature space. Our method is both simpler and more efficient than existing approaches to incremental learning, in terms of model size, storage, and computation: it requires only a single, fixed-capacity autoencoding network with a feature space that is used for both discriminative and generative purposes. All network parameters are optimized simultaneously without architectural manipulations, by solving a constrained minimax game between the encoding and decoding maps over a single rate reduction-based objective. Experimental results show that our method can effectively alleviate catastrophic forgetting, achieving significantly better performance than prior work for both generative and discriminative purposes.
14.Coded ResNeXt: a network for designing disentangled information paths ⬇️
To avoid treating neural networks as highly complex black boxes, the deep learning research community has tried to build interpretable models allowing humans to understand the decisions taken by the model. Unfortunately, the focus is mostly on manipulating only the very high-level features associated with the last layers. In this work, we look at neural network architectures for classification in a more general way and introduce an algorithm which defines before the training the paths of the network through which the per-class information flows. We show that using our algorithm we can extract a lighter single-purpose binary classifier for a particular class by removing the parameters that do not participate in the predefined information path of that class, which is approximately 60% of the total parameters. Notably, leveraging coding theory to design the information paths enables us to use intermediate network layers for making early predictions without having to evaluate the full network. We demonstrate that a slightly modified ResNeXt model, trained with our algorithm, can achieve higher classification accuracy on CIFAR-10/100 and ImageNet than the original ResNeXt, while having all the aforementioned properties.
15.Learning the Pedestrian-Vehicle Interaction for Pedestrian Trajectory Prediction ⬇️
In this paper, we study the interaction between pedestrians and vehicles and propose a novel neural network structure called the Pedestrian-Vehicle Interaction (PVI) extractor for learning the pedestrian-vehicle interaction. We implement the proposed PVI extractor on both sequential approaches (long short-term memory (LSTM) models) and non-sequential approaches (convolutional models). We use the Waymo Open Dataset that contains real-world urban traffic scenes with both pedestrian and vehicle annotations. For the LSTM-based models, our proposed model is compared with Social-LSTM and Social-GAN, and using our proposed PVI extractor reduces the average displacement error (ADE) and the final displacement error (FDE) by 7.46% and 5.24%, respectively. For the convolutional-based models, our proposed model is compared with Social-STGCNN and Social-IWSTCNN, and using our proposed PVI extractor reduces the ADE and FDE by 2.10% and 1.27%, respectively. The results show that the pedestrian-vehicle interaction influences pedestrian behavior, and the models using the proposed PVI extractor can capture the interaction between pedestrians and vehicles, and thereby outperform the compared methods.
16.Describing image focused in cognitive and visual details for visually impaired people: An approach to generating inclusive paragraphs ⬇️
Several services for people with visual disabilities have emerged recently due to achievements in Assistive Technologies and Artificial Intelligence areas. Despite the growth in assistive systems availability, there is a lack of services that support specific tasks, such as understanding the image context presented in online content, e.g., webinars. Image captioning techniques and their variants are limited as Assistive Technologies as they do not match the needs of visually impaired people when generating specific descriptions. We propose an approach for generating context of webinar images combining a dense captioning technique with a set of filters, to fit the captions in our domain, and a language model for the abstractive summary task. The results demonstrated that we can produce descriptions with higher interpretability and focused on the relevant information for that group of people by combining image analysis methods and neural language models.
17.Face Beneath the Ink: Synthetic Data and Tattoo Removal with Application to Face Recognition ⬇️
Systems that analyse faces have seen significant improvements in recent years and are today used in numerous application scenarios. However, these systems have been found to be negatively affected by facial alterations such as tattoos. To better understand and mitigate the effect of facial tattoos in facial analysis systems, large datasets of images of individuals with and without tattoos are needed. To this end, we propose a generator for automatically adding realistic tattoos to facial images. Moreover, we demonstrate the feasibility of the generation by training a deep learning-based model for removing tattoos from face images. The experimental results show that it is possible to remove facial tattoos from real images without degrading the quality of the image. Additionally, we show that it is possible to improve face recognition accuracy by using the proposed deep learning-based tattoo removal before extracting and comparing facial features.
18.A Field of Experts Prior for Adapting Neural Networks at Test Time ⬇️
Performance of convolutional neural networks (CNNs) in image analysis tasks is often marred in the presence of acquisition-related distribution shifts between training and test images. Recently, it has been proposed to tackle this problem by fine-tuning trained CNNs for each test image. Such test-time-adaptation (TTA) is a promising and practical strategy for improving robustness to distribution shifts as it requires neither data sharing between institutions nor annotating additional data. Previous TTA methods use a helper model to increase similarity between outputs and/or features extracted from a test image with those of the training images. Such helpers, which are typically modeled using CNNs, can be task-specific and themselves vulnerable to distribution shifts in their inputs. To overcome these problems, we propose to carry out TTA by matching the feature distributions of test and training images, as modelled by a field-of-experts (FoE) prior. FoEs model complicated probability distributions as products of many simpler expert distributions. We use 1D marginal distributions of a trained task CNN's features as experts in the FoE model. Further, we compute principal components of patches of the task CNN's features, and consider the distributions of PCA loadings as additional experts. We validate the method on 5 MRI segmentation tasks (healthy tissues in 4 anatomical regions and lesions in 1 one anatomy), using data from 17 clinics, and on a MRI registration task, using data from 3 clinics. We find that the proposed FoE-based TTA is generically applicable in multiple tasks, and outperforms all previous TTA methods for lesion segmentation. For healthy tissue segmentation, the proposed method outperforms other task-agnostic methods, but a previous TTA method which is specifically designed for segmentation performs the best for most of the tested datasets. Our code is publicly available.
19.SafePicking: Learning Safe Object Extraction via Object-Level Mapping ⬇️
Robots need object-level scene understanding to manipulate objects while reasoning about contact, support, and occlusion among objects. Given a pile of objects, object recognition and reconstruction can identify the boundary of object instances, giving important cues as to how the objects form and support the pile. In this work, we present a system, SafePicking, that integrates object-level mapping and learning-based motion planning to generate a motion that safely extracts occluded target objects from a pile. Planning is done by learning a deep Q-network that receives observations of predicted poses and a depth-based heightmap to output a motion trajectory, trained to maximize a safety metric reward. Our results show that the observation fusion of poses and depth-sensing gives both better performance and robustness to the model. We evaluate our methods using the YCB objects in both simulation and the real world, achieving safe object extraction from piles.
20.CLIPasso: Semantically-Aware Object Sketching ⬇️
Abstraction is at the heart of sketching due to the simple and minimal nature of line drawings. Abstraction entails identifying the essential visual properties of an object or scene, which requires semantic understanding and prior knowledge of high-level concepts. Abstract depictions are therefore challenging for artists, and even more so for machines. We present an object sketching method that can achieve different levels of abstraction, guided by geometric and semantic simplifications. While sketch generation methods often rely on explicit sketch datasets for training, we utilize the remarkable ability of CLIP (Contrastive-Language-Image-Pretraining) to distill semantic concepts from sketches and images alike. We define a sketch as a set of Bézier curves and use a differentiable rasterizer to optimize the parameters of the curves directly with respect to a CLIP-based perceptual loss. The abstraction degree is controlled by varying the number of strokes. The generated sketches demonstrate multiple levels of abstraction while maintaining recognizability, underlying structure, and essential visual components of the subject drawn.
21.Meta-learning with GANs for anomaly detection, with deployment in high-speed rail inspection system ⬇️
Anomaly detection has been an active research area with a wide range of potential applications. Key challenges for anomaly detection in the AI era with big data include lack of prior knowledge of potential anomaly types, highly complex and noisy background in input data, scarce abnormal samples, and imbalanced training dataset. In this work, we propose a meta-learning framework for anomaly detection to deal with these issues. Within this framework, we incorporate the idea of generative adversarial networks (GANs) with appropriate choices of loss functions including structural similarity index measure (SSIM). Experiments with limited labeled data for high-speed rail inspection demonstrate that our meta-learning framework is sharp and robust in identifying anomalies. Our framework has been deployed in five high-speed railways of China since 2021: it has reduced more than 99.7% workload and saved 96.7% inspection time.
22.Multi-Modal Knowledge Graph Construction and Application: A Survey ⬇️
Recent years have witnessed the resurgence of knowledge engineering which is featured by the fast growth of knowledge graphs. However, most of existing knowledge graphs are represented with pure symbols, which hurts the machine's capability to understand the real world. The multi-modalization of knowledge graphs is an inevitable key step towards the realization of human-level machine intelligence. The results of this endeavor are Multi-modal Knowledge Graphs (MMKGs). In this survey on MMKGs constructed by texts and images, we first give definitions of MMKGs, followed with the preliminaries on multi-modal tasks and techniques. We then systematically review the challenges, progresses and opportunities on the construction and application of MMKGs respectively, with detailed analyses of the strength and weakness of different solutions. We finalize this survey with open research problems relevant to MMKGs.
23.Assessing Privacy Risks from Feature Vector Reconstruction Attacks ⬇️
In deep neural networks for facial recognition, feature vectors are numerical representations that capture the unique features of a given face. While it is known that a version of the original face can be recovered via "feature reconstruction," we lack an understanding of the end-to-end privacy risks produced by these attacks. In this work, we address this shortcoming by developing metrics that meaningfully capture the threat of reconstructed face images. Using end-to-end experiments and user studies, we show that reconstructed face images enable re-identification by both commercial facial recognition systems and humans, at a rate that is at worst, a factor of four times higher than randomized baselines. Our results confirm that feature vectors should be recognized as Personal Identifiable Information (PII) in order to protect user privacy.
24.Towards Adversarially Robust Deepfake Detection: An Ensemble Approach ⬇️
Detecting deepfakes is an important problem, but recent work has shown that DNN-based deepfake detectors are brittle against adversarial deepfakes, in which an adversary adds imperceptible perturbations to a deepfake to evade detection. In this work, we show that a modification to the detection strategy in which we replace a single classifier with a carefully chosen ensemble, in which input transformations for each model in the ensemble induces pairwise orthogonal gradients, can significantly improve robustness beyond the de facto solution of adversarial training. We present theoretical results to show that such orthogonal gradients can help thwart a first-order adversary by reducing the dimensionality of the input subspace in which adversarial deepfakes lie. We validate the results empirically by instantiating and evaluating a randomized version of such "orthogonal" ensembles for adversarial deepfake detection and find that these randomized ensembles exhibit significantly higher robustness as deepfake detectors compared to state-of-the-art deepfake detectors against adversarial deepfakes, even those created using strong PGD-500 attacks.
25.Vehicle and License Plate Recognition with Novel Dataset for Toll Collection ⬇️
We propose an automatic framework for toll collection, consisting of three steps: vehicle type recognition, license plate localization, and reading. However, each of the three steps becomes non-trivial due to image variations caused by several factors. The traditional vehicle decorations on the front cause variations among vehicles of the same type. These decorations make license plate localization and recognition difficult due to severe background clutter and partial occlusions. Likewise, on most vehicles, specifically trucks, the position of the license plate is not consistent. Lastly, for license plate reading, the variations are induced by non-uniform font styles, sizes, and partially occluded letters and numbers. Our proposed framework takes advantage of both data availability and performance evaluation of the backbone deep learning architectures. We gather a novel dataset, \emph{Diverse Vehicle and License Plates Dataset (DVLPD)}, consisting of 10k images belonging to six vehicle types. Each image is then manually annotated for vehicle type, license plate, and its characters and digits. For each of the three tasks, we evaluate You Only Look Once (YOLO)v2, YOLOv3, YOLOv4, and FasterRCNN. For real-time implementation on a Raspberry Pi, we evaluate the lighter versions of YOLO named Tiny YOLOv3 and Tiny YOLOv4. The best Mean Average Precision ([email protected]) of 98.8% for vehicle type recognition, 98.5% for license plate detection, and 98.3% for license plate reading is achieved by YOLOv4, while its lighter version, i.e., Tiny YOLOv4 obtained a mAP of 97.1%, 97.4%, and 93.7% on vehicle type recognition, license plate detection, and license plate reading, respectively. The dataset and the training codes are available at this https URL
26.Artemis: Articulated Neural Pets with Appearance and Motion synthesis ⬇️
We human are entering into a virtual era, and surely want to bring animals to virtual world as well for companion. Yet, computer-generated (CGI) furry animals is limited by tedious off-line rendering, let alone interactive motion control. In this paper, we present ARTEMIS, a novel neural modeling and rendering pipeline for generating ARTiculated neural pets with appEarance and Motion synthesIS. Our ARTEMIS enables interactive motion control, real-time animation and photo-realistic rendering of furry animals. The core of ARTEMIS is a neural-generated (NGI) animal engine, which adopts an efficient octree based representation for animal animation and fur rendering. The animation then becomes equivalent to voxel level skeleton based deformation. We further use a fast octree indexing, an efficient volumetric rendering scheme to generate appearance and density features maps. Finally, we propose a novel shading network to generate high-fidelity details of appearance and opacity under novel poses. For the motion control module in ARTEMIS, we combine state-of-the-art animal motion capture approach with neural character control scheme. We introduce an effective optimization scheme to reconstruct skeletal motion of real animals captured by a multi-view RGB and Vicon camera array. We feed the captured motion into a neural character control scheme to generate abstract control signals with motion styles. We further integrate ARTEMIS into existing engines that support VR headsets, providing an unprecedented immersive experience where a user can intimately interact with a variety of virtual animals with vivid movements and photo-realistic appearance. Extensive experiments and showcases demonstrate the effectiveness of our ARTEMIS system to achieve highly realistic rendering of NGI animals in real-time, providing daily immersive and interactive experience with digital animals unseen before.
27.A Wasserstein GAN for Joint Learning of Inpainting and its Spatial Optimisation ⬇️
Classic image inpainting is a restoration method that reconstructs missing image parts. However, a carefully selected mask of known pixels that yield a high quality inpainting can also act as a sparse image representation. This challenging spatial optimisation problem is essential for practical applications such as compression. So far, it has been almost exclusively addressed by model-based approaches. First attempts with neural networks seem promising, but are tailored towards specific inpainting operators or require postprocessing. To address this issue, we propose the first generative adversarial network for spatial inpainting data optimisation. In contrast to previous approaches, it allows joint training of an inpainting generator and a corresponding mask optimisation network. With a Wasserstein distance, we ensure that our inpainting results accurately reflect the statistics of natural images. This yields significant improvements in visual quality and speed over conventional stochastic models and also outperforms current spatial optimisation networks.
28.Unsupervised HDR Imaging: What Can Be Learned from a Single 8-bit Video? ⬇️
Recently, Deep Learning-based methods for inverse tone-mapping standard dynamic range (SDR) images to obtain high dynamic range (HDR) images have become very popular. These methods manage to fill over-exposed areas convincingly both in terms of details and dynamic range. Typically, these methods, to be effective, need to learn from large datasets and to transfer this knowledge to the network weights. In this work, we tackle this problem from a completely different perspective. What can we learn from a single SDR video? With the presented zero-shot approach, we show that, in many cases, a single SDR video is sufficient to be able to generate an HDR video of the same quality or better than other state-of-the-art methods.
29.Dilated convolutional neural network-based deep reference picture generation for video compression ⬇️
Motion estimation and motion compensation are indispensable parts of inter prediction in video coding. Since the motion vector of objects is mostly in fractional pixel units, original reference pictures may not accurately provide a suitable reference for motion compensation. In this paper, we propose a deep reference picture generator which can create a picture that is more relevant to the current encoding frame, thereby further reducing temporal redundancy and improving video compression efficiency. Inspired by the recent progress of Convolutional Neural Network(CNN), this paper proposes to use a dilated CNN to build the generator. Moreover, we insert the generated deep picture into Versatile Video Coding(VVC) as a reference picture and perform a comprehensive set of experiments to evaluate the effectiveness of our network on the latest VVC Test Model VTM. The experimental results demonstrate that our proposed method achieves on average 9.7% bit saving compared with VVC under low-delay P configuration.
30.Entroformer: A Transformer-based Entropy Model for Learned Image Compression ⬇️
One critical component in lossy deep image compression is the entropy model, which predicts the probability distribution of the quantized latent representation in the encoding and decoding modules. Previous works build entropy models upon convolutional neural networks which are inefficient in capturing global dependencies. In this work, we propose a novel transformer-based entropy model, termed Entroformer, to capture long-range dependencies in probability distribution estimation effectively and efficiently. Different from vision transformers in image classification, the Entroformer is highly optimized for image compression, including a top-k self-attention and a diamond relative position encoding. Meanwhile, we further expand this architecture with a parallel bidirectional context model to speed up the decoding process. The experiments show that the Entroformer achieves state-of-the-art performance on image compression while being time-efficient.
31.Including Facial Expressions in Contextual Embeddings for Sign Language Generation ⬇️
State-of-the-art sign language generation frameworks lack expressivity and naturalness which is the result of only focusing manual signs, neglecting the affective, grammatical and semantic functions of facial expressions. The purpose of this work is to augment semantic representation of sign language through grounding facial expressions. We study the effect of modeling the relationship between text, gloss, and facial expressions on the performance of the sign generation systems. In particular, we propose a Dual Encoder Transformer able to generate manual signs as well as facial expressions by capturing the similarities and differences found in text and sign gloss annotation. We take into consideration the role of facial muscle activity to express intensities of manual signs by being the first to employ facial action units in sign language generation. We perform a series of experiments showing that our proposed model improves the quality of automatically generated sign language.
32.Give me a knee radiograph, I will tell you where the knee joint area is: a deep convolutional neural network adventure ⬇️
Knee pain is undoubtedly the most common musculoskeletal symptom that impairs quality of life, confines mobility and functionality across all ages. Knee pain is clinically evaluated by routine radiographs, where the widespread adoption of radiographic images and their availability at low cost, make them the principle component in the assessment of knee pain and knee pathologies, such as arthritis, trauma, and sport injuries. However, interpretation of the knee radiographs is still highly subjective, and overlapping structures within the radiographs and the large volume of images needing to be analyzed on a daily basis, make interpretation challenging for both naive and experienced practitioners. There is thus a need to implement an artificial intelligence strategy to objectively and automatically interpret knee radiographs, facilitating triage of abnormal radiographs in a timely fashion. The current work proposes an accurate and effective pipeline for autonomous detection, localization, and classification of knee joint area in plain radiographs combining the You Only Look Once (YOLO v3) deep convolutional neural network with a large and fully-annotated knee radiographs dataset. The present work is expected to stimulate more interest from the deep learning computer vision community to this pragmatic and clinical application.
33.The MeLa BitChute Dataset ⬇️
In this paper we present a near-complete dataset of over 3M videos from 61K channels over 2.5 years (June 2019 to December 2021) from the social video hosting platform BitChute, a commonly used alternative to YouTube. Additionally, we include a variety of video-level metadata, including comments, channel descriptions, and views for each video. The MeLa-BitChute dataset can be found at: this https URL.
34.Optimal Transport for Super Resolution Applied to Astronomy Imaging ⬇️
Super resolution is an essential tool in optics, especially on interstellar scales, due to physical laws restricting possible imaging resolution. We propose using optimal transport and entropy for super resolution applications. We prove that the reconstruction is accurate when sparsity is known and noise or distortion is small enough. We prove that the optimizer is stable and robust to noise and perturbations. We compare this method to a state of the art convolutional neural network and get similar results for much less computational cost and greater methodological flexibility.
35.Domain Adversarial Training: A Game Perspective ⬇️
The dominant line of work in domain adaptation has focused on learning invariant representations using domain-adversarial training. In this paper, we interpret this approach from a game theoretical perspective. Defining optimal solutions in domain-adversarial training as a local Nash equilibrium, we show that gradient descent in domain-adversarial training can violate the asymptotic convergence guarantees of the optimizer, oftentimes hindering the transfer performance. Our analysis leads us to replace gradient descent with high-order ODE solvers (i.e., Runge-Kutta), for which we derive asymptotic convergence guarantees. This family of optimizers is significantly more stable and allows more aggressive learning rates, leading to high performance gains when used as a drop-in replacement over standard optimizers. Our experiments show that in conjunction with state-of-the-art domain-adversarial methods, we achieve up to 3.5% improvement with less than of half training iterations. Our optimizers are easy to implement, free of additional parameters, and can be plugged into any domain-adversarial framework.
36.Dynamic Background Subtraction by Generative Neural Networks ⬇️
Background subtraction is a significant task in computer vision and an essential step for many real world applications. One of the challenges for background subtraction methods is dynamic background, which constitute stochastic movements in some parts of the background. In this paper, we have proposed a new background subtraction method, called DBSGen, which uses two generative neural networks, one for dynamic motion removal and another for background generation. At the end, the foreground moving objects are obtained by a pixel-wise distance threshold based on a dynamic entropy map. The proposed method has a unified framework that can be optimized in an end-to-end and unsupervised fashion. The performance of the method is evaluated over dynamic background sequences and it outperforms most of state-of-the-art methods. Our code is publicly available at this https URL.
37.Mining the manifolds of deep generative models for multiple data-consistent solutions of ill-posed tomographic imaging problems ⬇️
Tomographic imaging is in general an ill-posed inverse problem. Typically, a single regularized image estimate of the sought-after object is obtained from tomographic measurements. However, there may be multiple objects that are all consistent with the same measurement data. The ability to generate such alternate solutions is important because it may enable new assessments of imaging systems. In principle, this can be achieved by means of posterior sampling methods. In recent years, deep neural networks have been employed for posterior sampling with promising results. However, such methods are not yet for use with large-scale tomographic imaging applications. On the other hand, empirical sampling methods may be computationally feasible for large-scale imaging systems and enable uncertainty quantification for practical applications. Empirical sampling involves solving a regularized inverse problem within a stochastic optimization framework in order to obtain alternate data-consistent solutions. In this work, we propose a new empirical sampling method that computes multiple solutions of a tomographic inverse problem that are consistent with the same acquired measurement data. The method operates by repeatedly solving an optimization problem in the latent space of a style-based generative adversarial network (StyleGAN), and was inspired by the Photo Upsampling via Latent Space Exploration (PULSE) method that was developed for super-resolution tasks. The proposed method is demonstrated and analyzed via numerical studies that involve two stylized tomographic imaging modalities. These studies establish the ability of the method to perform efficient empirical sampling and uncertainty quantification.
38.Characterizing and overcoming the greedy nature of learning in multi-modal deep neural networks ⬇️
We hypothesize that due to the greedy nature of learning in multi-modal deep neural networks, these models tend to rely on just one modality while under-fitting the other modalities. Such behavior is counter-intuitive and hurts the models' generalization, as we observe empirically. To estimate the model's dependence on each modality, we compute the gain on the accuracy when the model has access to it in addition to another modality. We refer to this gain as the conditional utilization rate. In the experiments, we consistently observe an imbalance in conditional utilization rates between modalities, across multiple tasks and architectures. Since conditional utilization rate cannot be computed efficiently during training, we introduce a proxy for it based on the pace at which the model learns from each modality, which we refer to as the conditional learning speed. We propose an algorithm to balance the conditional learning speeds between modalities during training and demonstrate that it indeed addresses the issue of greedy learning. The proposed algorithm improves the model's generalization on three datasets: Colored MNIST, Princeton ModelNet40, and NVIDIA Dynamic Hand Gesture.
39.Motion Puzzle: Arbitrary Motion Style Transfer by Body Part ⬇️
This paper presents Motion Puzzle, a novel motion style transfer network that advances the state-of-the-art in several important respects. The Motion Puzzle is the first that can control the motion style of individual body parts, allowing for local style editing and significantly increasing the range of stylized motions. Designed to keep the human's kinematic structure, our framework extracts style features from multiple style motions for different body parts and transfers them locally to the target body parts. Another major advantage is that it can transfer both global and local traits of motion style by integrating the adaptive instance normalization and attention modules while keeping the skeleton topology. Thus, it can capture styles exhibited by dynamic movements, such as flapping and staggering, significantly better than previous work. In addition, our framework allows for arbitrary motion style transfer without datasets with style labeling or motion pairing, making many publicly available motion datasets available for training. Our framework can be easily integrated with motion generation frameworks to create many applications, such as real-time motion transfer. We demonstrate the advantages of our framework with a number of examples and comparisons with previous work.
40.Towards a Guideline for Evaluation Metrics in Medical Image Segmentation ⬇️
In the last decade, research on artificial intelligence has seen rapid growth with deep learning models, especially in the field of medical image segmentation. Various studies demonstrated that these models have powerful prediction capabilities and achieved similar results as clinicians. However, recent studies revealed that the evaluation in image segmentation studies lacks reliable model performance assessment and showed statistical bias by incorrect metric implementation or usage. Thus, this work provides an overview and interpretation guide on the following metrics for medical image segmentation evaluation in binary as well as multi-class problems: Dice similarity coefficient, Jaccard, Sensitivity, Specificity, Rand index, ROC curves, Cohen's Kappa, and Hausdorff distance. As a summary, we propose a guideline for standardized medical image segmentation evaluation to improve evaluation quality, reproducibility, and comparability in the research field.
41.A Deep Learning Approach for Digital ColorReconstruction of Lenticular Films ⬇️
We propose the first accurate digitization and color reconstruction process for historical lenticular film that is robust to artifacts. Lenticular films emerged in the 1920s and were one of the first technologies that permitted to capture full color information in motion. The technology leverages an RGB filter and cylindrical lenticules embossed on the film surface to encode the color in the horizontal spatial dimension of the image. To project the pictures the encoding process was reversed using an appropriate analog device. In this work, we introduce an automated, fully digital pipeline to process the scan of lenticular films and colorize the image. Our method merges deep learning with a model-based approach in order to maximize the performance while making sure that the reconstructed colored images truthfully match the encoded color information. Our model employs different strategies to achieve an effective color reconstruction, in particular (i) we use data augmentation to create a robust lenticule segmentation network, (ii) we fit the lenticules raster prediction to obtain a precise vectorial lenticule localization, and (iii) we train a colorization network that predicts interpolation coefficients in order to obtain a truthful colorization. We validate the proposed method on a lenticular film dataset and compare it to other approaches. Since no colored groundtruth is available as reference, we conduct a user study to validate our method in a subjective manner. The results of the study show that the proposed method is largely preferred with respect to other existing and baseline methods.
42.A Plug-and-Play Approach to Multiparametric Quantitative MRI: Image Reconstruction using Pre-Trained Deep Denoisers ⬇️
Current spatiotemporal deep learning approaches to Magnetic Resonance Fingerprinting (MRF) build artefact-removal models customised to a particular k-space subsampling pattern which is used for fast (compressed) acquisition. This may not be useful when the acquisition process is unknown during training of the deep learning model and/or changes during testing time. This paper proposes an iterative deep learning plug-and-play reconstruction approach to MRF which is adaptive to the forward acquisition process. Spatiotemporal image priors are learned by an image denoiser i.e. a Convolutional Neural Network (CNN), trained to remove generic white gaussian noise (not a particular subsampling artefact) from data. This CNN denoiser is then used as a data-driven shrinkage operator within the iterative reconstruction algorithm. This algorithm with the same denoiser model is then tested on two simulated acquisition processes with distinct subsampling patterns. The results show consistent de-aliasing performance against both acquisition schemes and accurate mapping of tissues' quantitative bio-properties. Software available: this https URL
43.HNF-Netv2 for Brain Tumor Segmentation using multi-modal MR Imaging ⬇️
In our previous work,
$i.e.$ , HNF-Net, high-resolution feature representation and light-weight non-local self-attention mechanism are exploited for brain tumor segmentation using multi-modal MR imaging. In this paper, we extend our HNF-Net to HNF-Netv2 by adding inter-scale and intra-scale semantic discrimination enhancing blocks to further exploit global semantic discrimination for the obtained high-resolution features. We trained and evaluated our HNF-Netv2 on the multi-modal Brain Tumor Segmentation Challenge (BraTS) 2021 dataset. The result on the test set shows that our HNF-Netv2 achieved the average Dice scores of 0.878514, 0.872985, and 0.924919, as well as the Hausdorff distances ($95%$ ) of 8.9184, 16.2530, and 4.4895 for the enhancing tumor, tumor core, and whole tumor, respectively. Our method won the RSNA 2021 Brain Tumor AI Challenge Prize (Segmentation Task), which ranks 8th out of all 1250 submitted results.
44.On Real-time Image Reconstruction with Neural Networks for MRI-guided Radiotherapy ⬇️
MRI-guidance techniques that dynamically adapt radiation beams to follow tumor motion in real-time will lead to more accurate cancer treatments and reduced collateral healthy tissue damage. The gold-standard for reconstruction of undersampled MR data is compressed sensing (CS) which is computationally slow and limits the rate that images can be available for real-time adaptation. Here, we demonstrate the use of automated transform by manifold approximation (AUTOMAP), a generalized framework that maps raw MR signal to the target image domain, to rapidly reconstruct images from undersampled radial k-space data. The AUTOMAP neural network was trained to reconstruct images from a golden-angle radial acquisition, a benchmark for motion-sensitive imaging, on lung cancer patient data and generic images from ImageNet. Model training was subsequently augmented with motion-encoded k-space data derived from videos in the YouTube-8M dataset to encourage motion robust reconstruction. We find that AUTOMAP-reconstructed radial k-space has equivalent accuracy to CS but with much shorter processing times after initial fine-tuning on retrospectively acquired lung cancer patient data. Validation of motion-trained models with a virtual dynamic lung tumor phantom showed that the generalized motion properties learned from YouTube lead to improved target tracking accuracy. Our work shows that AUTOMAP can achieve real-time, accurate reconstruction of radial data. These findings imply that neural-network-based reconstruction is potentially superior to existing approaches for real-time image guidance applications.