Abstract: Humans exploit prior knowledge to describe images, and are able to adapt their explanation to specific contextual information given, even to the extent of inventing plausible explanations when contextual information and images do not match. In this work, we propose the novel task of captioning Wikipedia images by integrating contextual knowledge. Specifically, we produce models that jointly reason over Wikipedia articles, Wikimedia images and their associated descriptions to produce contextualized captions. The same Wikimedia image can be used to illustrate different articles, and the produced caption needs to be adapted to the specific context allowing us to explore the limits of the model to adjust captions to different contextual information. Dealing with out-of-dictionary words and Named Entities is a challenging task in this domain. To address this, we propose a pre-training objective, Masked Named Entity Modeling (MNEM), and show that this pretext task results to significantly improved models. Furthermore, we verify that a model pre-trained in Wikipedia generalizes well to News Captioning datasets. We further define two different test splits according to the difficulty of the captioning task. We offer insights on the role and the importance of each modality and highlight the limitations of our model.

Abstract: Transformer models have shown great success handling long-range interactions, making them a promising tool for modeling video. However, they lack inductive biases and scale quadratically with input length. These limitations are further exacerbated when dealing with the high dimensionality introduced by the temporal dimension. While there are surveys analyzing the advances of Transformers for vision, none focus on an in-depth analysis of video-specific designs. In this survey, we analyze the main contributions and trends of works leveraging Transformers to model video. Specifically, we delve into how videos are handled at the input level first. Then, we study the architectural changes made to deal with video more efficiently, reduce redundancy, re-introduce useful inductive biases, and capture long-term temporal dynamics. In addition, we provide an overview of different training regimes and explore effective self-supervised learning strategies for video. Finally, we conduct a performance comparison on the most common benchmark for Video Transformers (i.e., action classification), finding them to outperform 3D ConvNets even with less computational complexity.

Abstract: Domain adaptation (DA) aims to alleviate the domain shift between source domain and target domain. Most DA methods require access to the source data, but often that is not possible (e.g., due to data privacy or intellectual property). In this paper, we address the challenging source-free domain adaptation (SFDA) problem, where the source pretrained model is adapted to the target domain in the absence of source data. Our method is based on the observation that target data, which might not align with the source domain classifier, still forms clear clusters. We capture this intrinsic structure by defining local affinity of the target data, and encourage label consistency among data with high local affinity. We observe that higher affinity should be assigned to reciprocal neighbors. To aggregate information with more context, we consider expanded neighborhoods with small affinity values. Furthermore, we consider the density around each target sample, which can alleviate the negative impact of potential outliers. In the experimental results we verify that the inherent structure of the target features is an important source of information for domain adaptation. We demonstrate that this local structure can be efficiently captured by considering the local neighbors, the reciprocal neighbors, and the expanded neighborhood. Finally, we achieve state-of-the-art performance on several 2D image and 3D point cloud recognition datasets.

Abstract: Gesture Recognition (GR) is a challenging research area in computer vision. To tackle the annotation bottleneck in GR, we formulate the problem of Zero-Shot Gesture Recognition (ZS-GR) and propose a two-stream model from two input modalities: RGB and Depth videos. To benefit from the vision Transformer capabilities, we use two vision Transformer models, for human detection and visual features representation. We configure a transformer encoder-decoder architecture, as a fast and accurate human detection model, to overcome the challenges of the current human detection models. Considering the human keypoints, the detected human body is segmented into nine parts. A spatio-temporal representation from human body is obtained using a vision Transformer and a LSTM network. A semantic space maps the visual features to the lingual embedding of the class labels via a Bidirectional Encoder Representations from Transformers (BERT) model. We evaluated the proposed model on five datasets, Montalbano II, MSR Daily Activity 3D, CAD-60, NTU-60, and isoGD obtaining state-of-the-art results compared to state-of-the-art ZS-GR models as well as the Zero-Shot Action Recognition (ZS-AR).

Abstract: Domain Adaptation (DA) has recently raised strong interests in the medical imaging community. While a large variety of DA techniques has been proposed for image segmentation, most of these techniques have been validated either on private datasets or on small publicly available datasets. Moreover, these datasets mostly addressed single-class problems. To tackle these limitations, the Cross-Modality Domain Adaptation (crossMoDA) challenge was organised in conjunction with the 24th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2021). CrossMoDA is the first large and multi-class benchmark for unsupervised cross-modality DA. The challenge's goal is to segment two key brain structures involved in the follow-up and treatment planning of vestibular schwannoma (VS): the VS and the cochleas. Currently, the diagnosis and surveillance in patients with VS are performed using contrast-enhanced T1 (ceT1) MRI. However, there is growing interest in using non-contrast sequences such as high-resolution T2 (hrT2) MRI. Therefore, we created an unsupervised cross-modality segmentation benchmark. The training set provides annotated ceT1 (N=105) and unpaired non-annotated hrT2 (N=105). The aim was to automatically perform unilateral VS and bilateral cochlea segmentation on hrT2 as provided in the testing set (N=137). A total of 16 teams submitted their algorithm for the evaluation phase. The level of performance reached by the top-performing teams is strikingly high (best median Dice – VS:88.4%; Cochleas:85.7%) and close to full supervision (median Dice – VS:92.5%; Cochleas:87.7%). All top-performing methods made use of an image-to-image translation approach to transform the source-domain images into pseudo-target-domain images. A segmentation network was then trained using these generated images and the manual annotations provided for the source image.

Abstract: Deep neural networks represent a compelling technique to tackle complex real-world problems, but are over-parameterized and often suffer from over- or under-confident estimates. Deep ensembles have shown better parameter estimations and often provide reliable uncertainty estimates that contribute to the robustness of the results. In this work, we propose a new metric to identify samples that are hard to classify. Our metric is defined as coincidence score for deep ensembles which measures the agreement of its individual models. The main hypothesis we rely on is that deep learning algorithms learn the low-loss samples better compared to large-loss samples. In order to compensate for this, we use controlled over-sampling on the identified ”hard” samples using proper data augmentation schemes to enable the models to learn those samples better. We validate the proposed metric using two public food datasets on different backbone architectures and show the improvements compared to the conventional deep neural network training using different performance metrics.

Abstract: Recent advances in single image super-resolution (SISR) explore the power of deep convolutional neural networks (CNNs) to achieve better performance. However, most of the progress has been made by scaling CNN architectures, which usually raise computational demands and memory consumption. This makes modern architectures less applicable in practice. In addition, most CNN-based SR methods do not fully utilize the informative hierarchical features that are helpful for final image recovery. In order to address these issues, we propose a directional variance attention network (DiVANet), a computationally efficient yet accurate network for SISR. Specifically, we introduce a novel directional variance attention (DiVA) mechanism to capture long-range spatial dependencies and exploit inter-channel dependencies simultaneously for more discriminative representations. Furthermore, we propose a residual attention feature group (RAFG) for parallelizing attention and residual block computation. The output of each residual block is linearly fused at the RAFG output to provide access to the whole feature hierarchy. In parallel, DiVA extracts most relevant features from the network for improving the final output and preventing information loss along the successive operations inside the network. Experimental results demonstrate the superiority of DiVANet over the state of the art in several datasets, while maintaining relatively low computation and memory footprint. The code is available at https://github.com/pbehjatii/DiVANet.