FBK-HUPBA Submission to the EPIC-Kitchens Action Recognition 2020 Challenge

In this report we describe the technical details of our submission to the EPIC-Kitchens Action Recognition 2020 Challenge. To participate in the challenge we deployed spatio-temporal feature extraction and aggregation models we have developed recently: Gate-Shift Module (GSM) [1] and EgoACO, an extension of Long Short-Term Attention (LSTA) [2]. We design an ensemble of GSM and EgoACO model families with different backbones and pre-training to generate the prediction scores. Our submission, visible on the public leaderboard with team name FBK-HUPBA, achieved a top-1 action recognition accuracy of 40.0% on S1 setting, and 25.71% on S2 setting, using only RGB.

Computing the Testing Error without a Testing Set

Deep Neural Networks (DNNs) have revolutionized computer vision. We now have DNNs that achieve top (performance) results in many problems, including object recognition, facial expression analysis, and semantic segmentation, to name but a few. The design of the DNNs that achieve top results is, however, non-trivial and mostly done by trail-and-error. That is, typically, researchers will derive many DNN architectures (i.e., topologies) and then test them on multiple datasets. However, there are no guarantees that the selected DNN will perform well in the real world. One can use a testing set to estimate the performance gap between the training and testing sets, but avoiding overfitting-to-the-testing-data is almost impossible. Using a sequestered testing dataset may address this problem, but this requires a constant update of the dataset, a very expensive venture. Here, we derive an algorithm to estimate the performance gap between training and testing that does not require any testing dataset. Specifically, we derive a number of persistent topology measures that identify when a DNN is learning to generalize to unseen samples. This allows us to compute the DNN's testing error on unseen samples, even when we do not have access to them. We provide extensive experimental validation on multiple networks and datasets to demonstrate the feasibility of the proposed approach.

Cross-ethnicity Face Anti-spoofing Recognition Challenge: A Review

Face anti-spoofing is critical to prevent face recognition systems from a security breach. The biometrics community has %possessed achieved impressive progress recently due the excellent performance of deep neural networks and the availability of large datasets. Although ethnic bias has been verified to severely affect the performance of face recognition systems, it still remains an open research problem in face anti-spoofing. Recently, a multi-ethnic face anti-spoofing dataset, CASIA-SURF CeFA, has been released with the goal of measuring the ethnic bias. It is the largest up to date cross-ethnicity face anti-spoofing dataset covering $3$ ethnicities, $3$ modalities, $1,607$ subjects, 2D plus 3D attack types, and the first dataset including explicit ethnic labels among the recently released datasets for face anti-spoofing. We organized the Chalearn Face Anti-spoofing Attack Detection Challenge which consists of single-modal (e.g., RGB) and multi-modal (e.g., RGB, Depth, Infrared (IR)) tracks around this novel resource to boost research aiming to alleviate the ethnic bias. Both tracks have attracted $340$ teams in the development stage, and finally 11 and 8 teams have submitted their codes in the single-modal and multi-modal face anti-spoofing recognition challenges, respectively. All the results were verified and re-ran by the organizing team, and the results were used for the final ranking. This paper presents an overview of the challenge, including its design, evaluation protocol and a summary of results. We analyze the top ranked solutions and draw conclusions derived from the competition. In addition we outline future work directions.

CASIA-SURF CeFA: A Benchmark for Multi-modal Cross-ethnicity Face Anti-spoofing

Ethnic bias has proven to negatively affect the performance of face recognition systems, and it remains an open research problem in face anti-spoofing. In order to study the ethnic bias for face anti-spoofing, we introduce the largest up to date CASIA-SURF Cross-ethnicity Face Anti-spoofing (CeFA) dataset (briefly named CeFA), covering $3$ ethnicities, $3$ modalities, $1,607$ subjects, and 2D plus 3D attack types. Four protocols are introduced to measure the affect under varied evaluation conditions, such as cross-ethnicity, unknown spoofs or both of them. To the best of our knowledge, CeFA is the first dataset including explicit ethnic labels in current published/released datasets for face anti-spoofing. Then, we propose a novel multi-modal fusion method as a strong baseline to alleviate these bias, namely, the static-dynamic fusion mechanism applied in each modality (i.e., RGB, Depth and infrared image). Later, a partially shared fusion strategy is proposed to learn complementary information from multiple modalities. Extensive experiments demonstrate that the proposed method achieves state-of-the-art results on the CASIA-SURF, OULU-NPU, SiW and the CeFA dataset.

Multi-level Context Gating of Embedded Collective Knowledge for Medical Image Segmentation

Medical image segmentation has been very challenging due to the large variation of anatomy across different cases. Recent advances in deep learning frameworks have exhibited faster and more accurate performance in image segmentation. Among the existing networks, U-Net has been successfully applied on medical image segmentation. In this paper, we propose an extension of U-Net for medical image segmentation, in which we take full advantages of U-Net, Squeeze and Excitation (SE) block, bi-directional ConvLSTM (BConvLSTM), and the mechanism of dense convolutions. (I) We improve the segmentation performance by utilizing SE modules within the U-Net, with a minor effect on model complexity. These blocks adaptively recalibrate the channel-wise feature responses by utilizing a self-gating mechanism of the global information embedding of the feature maps. (II) To strengthen feature propagation and encourage feature reuse, we use densely connected convolutions in the last convolutional layer of the encoding path. (III) Instead of a simple concatenation in the skip connection of U-Net, we employ BConvLSTM in all levels of the network to combine the feature maps extracted from the corresponding encoding path and the previous decoding up-convolutional layer in a non-linear way. The proposed model is evaluated on six datasets DRIVE, ISIC 2017 and 2018, lung segmentation, $PH^2$, and cell nuclei segmentation, achieving state-of-the-art performance.

Static and Dynamic Fusion for Multi-modal Cross-ethnicity Face Anti-spoofing

Regardless of the usage of deep learning and handcrafted methods, the dynamic information from videos and the effect of cross-ethnicity are rarely considered in face anti-spoofing. In this work, we propose a static-dynamic fusion mechanism for multi-modal face anti-spoofing. Inspired by motion divergences between real and fake faces, we incorporate the dynamic image calculated by rank pooling with static information into a conventional neural network (CNN) for each modality (i.e., RGB, Depth and infrared (IR)). Then, we develop a partially shared fusion method to learn complementary information from multiple modalities. Furthermore, in order to study the generalization capability of the proposal in terms of cross-ethnicity attacks and unknown spoofs, we introduce the largest public cross-ethnicity Face Anti-spoofing (CASIA-CeFA) dataset, covering 3 ethnicities, 3 modalities, 1607 subjects, and 2D plus 3D attack types. Experiments demonstrate that the proposed method achieves state-of-the-art results on CASIA-CeFA, CASIA-SURF, OULU-NPU and SiW.

CLOTH3D: Clothed 3D Humans

This work presents CLOTH3D, the first big scale synthetic dataset of 3D clothed human sequences. CLOTH3D contains a large variability on garment type, topology, shape, size, tightness and fabric. Clothes are simulated on top of thousands of different pose sequences and body shapes, generating realistic cloth dynamics. We provide the dataset with a generative model for cloth generation. We propose a Conditional Variational Auto-Encoder (CVAE) based on graph convolutions (GCVAE) to learn garment latent spaces. This allows for realistic generation of 3D garments on top of SMPL model for any pose and shape.

Gate-Shift Networks for Video Action Recognition

Deep 3D CNNs for video action recognition are designed to learn powerful representations in the joint spatio-temporal feature space. In practice however, because of the large number of parameters and computations involved, they may under-perform in the lack of sufficiently large datasets for training them at scale. In this paper we introduce spatial gating in spatial-temporal decomposition of 3D kernels. We implement this concept with Gate-Shift Module (GSM). GSM is lightweight and turns a 2D-CNN into a highly efficient spatio-temporal feature extractor. With GSM plugged in, a 2D-CNN learns to adaptively route features through time and combine them, at almost no additional parameters and computational overhead. We perform an extensive evaluation of the proposed module to study its effectiveness in video action recognition, achieving state-of-the-art results on Something Something-V1 and Diving48 datasets, and obtaining competitive results on EPIC-Kitchens with far less model complexity. With GSM plugged into TSN, on Something Something-V1 we obtain an absolute +32% boost in recognition accuracy (from 17.52% to 49.56%) with less than 1% additional parameters and computations. By ensembling models trained at different temporal scales, we reach beyond 55%.

On the Effect of Observed Subject Biases in Apparent Personality Analysis from Audio-visual Signals

Personality perception is implicitly biased due to many subjective factors, such as cultural, social, contextual, gender and appearance. Approaches developed for automatic personality perception are not expected to predict the real personality of the target, but the personality external observers attributed to it. Hence, they have to deal with human bias, inherently transferred to the training data. However, bias analysis in personality computing is an almost unexplored area. In this work, we study different possible sources of bias affecting personality perception, including emotions from facial expressions, attractiveness, age, gender, and ethnicity, as well as their influence on prediction ability for apparent personality estimation. To this end, we propose a multi-modal deep neural network that combines raw audio and visual information alongside predictions of attribute-specific models to regress apparent personality. We also analyse spatio-temporal aggregation schemes and the effect of different time intervals on first impressions. We base our study on the ChaLearn First Impressions dataset, consisting of one-person conversational videos. Our model shows state-of-the-art results regressing apparent personality based on the Big-Five model. Furthermore, given the interpretability nature of our network design, we provide an incremental analysis on the impact of each possible source of bias on final network predictions.