This report describes a multi-modal multi-task ($M^3$T) approach underlying our submission to the valence-arousal estimation track of the Affective Behavior Analysis in-the-wild (ABAW) Challenge, held in conjunction with the IEEE International Conference on Automatic Face and Gesture Recognition (FG) 2020. In the proposed $M^3$T framework, we fuse both visual features from videos and acoustic features from the audio tracks to estimate the valence and arousal. The spatio-temporal visual features are extracted with a 3D convolutional network and a bidirectional recurrent neural network. Considering the correlations between valence / arousal, emotions, and facial actions, we also explores mechanisms to benefit from other tasks. We evaluated the $M^3$T framework on the validation set provided by ABAW and it significantly outperforms the baseline method.
Retrieving videos of a particular person with face image as a query via hashing technique has many important applications. While face images are typically represented as vectors in Euclidean space, characterizing face videos with some robust set modeling techniques (e.g. covariance matrices as exploited in this study, which reside on Riemannian manifold), has recently shown appealing advantages. This hence results in a thorny heterogeneous spaces matching problem. Moreover, hashing with handcrafted features as done in many existing works is clearly inadequate to achieve desirable performance for this task. To address such problems, we present an end-to-end Deep Heterogeneous Hashing (DHH) method that integrates three stages including image feature learning, video modeling, and heterogeneous hashing in a single framework, to learn unified binary codes for both face images and videos. To tackle the key challenge of hashing on the manifold, a well-studied Riemannian kernel mapping is employed to project data (i.e. covariance matrices) into Euclidean space and thus enables to embed the two heterogeneous representations into a common Hamming space, where both intra-space discriminability and inter-space compatibility are considered. To perform network optimization, the gradient of the kernel mapping is innovatively derived via structured matrix backpropagation in a theoretically principled way. Experiments on three challenging datasets show that our method achieves quite competitive performance compared with existing hashing methods.
Combined variations containing low-resolution and occlusion often present in face images in the wild, e.g., under the scenario of video surveillance. While most of the existing face image recovery approaches can handle only one type of variation per model, in this work, we propose a deep generative adversarial network (FCSR-GAN) for performing joint face completion and face super-resolution via multi-task learning. The generator of FCSR-GAN aims to recover a high-resolution face image without occlusion given an input low-resolution face image with occlusion. The discriminator of FCSR-GAN uses a set of carefully designed losses (an adversarial loss, a perceptual loss, a pixel loss, a smooth loss, a style loss, and a face prior loss) to assure the high quality of the recovered high-resolution face images without occlusion. The whole network of FCSR-GAN can be trained end-to-end using our two-stage training strategy. Experimental results on the public-domain CelebA and Helen databases show that the proposed approach outperforms the state-of-the-art methods in jointly performing face super-resolution (up to 8 $\times$) and face completion, and shows good generalization ability in cross-database testing. Our FCSR-GAN is also useful for improving face identification performance when there are low-resolution and occlusion in face images.
Heart rate (HR) is an important physiological signal that reflects the physical and emotional status of a person. Traditional HR measurements usually rely on contact monitors, which may cause inconvenience and discomfort. Recently, some methods have been proposed for remote HR estimation from face videos; however, most of them focus on well-controlled scenarios, their generalization ability into less-constrained scenarios (e.g., with head movement, and bad illumination) are not known. At the same time, lacking large-scale HR databases has limited the use of deep models for remote HR estimation. In this paper, we propose an end-to-end RhythmNet for remote HR estimation from the face. In RyhthmNet, we use a spatial-temporal representation encoding the HR signals from multiple ROI volumes as its input. Then the spatial-temporal representations are fed into a convolutional network for HR estimation. We also take into account the relationship of adjacent HR measurements from a video sequence via Gated Recurrent Unit (GRU) and achieves efficient HR measurement. In addition, we build a large-scale multi-modal HR database (named as VIPL-HR, available at 'http://vipl.ict.ac.cn/view_database.php?id=15'), which contains 2,378 visible light videos (VIS) and 752 near-infrared (NIR) videos of 107 subjects. Our VIPL-HR database contains various variations such as head movements, illumination variations, and acquisition device changes, replicating a less-constrained scenario for HR estimation. The proposed approach outperforms the state-of-the-art methods on both the public-domain and our VIPL-HR databases.
Facial action units (AUs) recognition is essential for emotion analysis and has been widely applied in mental state analysis. Existing work on AU recognition usually requires big face dataset with AU labels; however, manual AU annotation requires expertise and can be time-consuming. In this work, we propose a semi-supervised approach for AU recognition utilizing a large number of web face images without AU labels and a relatively small face dataset with AU annotations inspired by the co-training methods. Unlike traditional co-training methods that require provided multi-view features and model re-training, we propose a novel co-training method, namely multi-label co-regularization, for semi-supervised facial AU recognition. Two deep neural networks are utilized to generate multi-view features for both labeled and unlabeled face images, and a multi-view loss is designed to enforce the two feature generators to get conditional independent representations. In order to constrain the prediction consistency of the two views, we further propose a multi-label co-regularization loss by minimizing the distance of the predicted AU probability distributions of two views. In addition, prior knowledge of the relationship between individual AUs is embedded through a graph convolutional network (GCN) for exploiting useful information from the big unlabeled dataset. Experiments on several benchmarks show that the proposed approach can effectively leverage large datasets of face images without AU labels to improve the AU recognition accuracy and outperform the state-of-the-art semi-supervised AU recognition methods.
Few-shot classification aims to recognize unlabeled samples from unseen classes given only few labeled samples. The unseen classes and low-data problem make few-shot classification very challenging. Many existing approaches extracted features from labeled and unlabeled samples independently, as a result, the features are not discriminative enough. In this work, we propose a novel Cross Attention Network to address the challenging problems in few-shot classification. Firstly, Cross Attention Module is introduced to deal with the problem of unseen classes. The module generates cross attention maps for each pair of class feature and query sample feature so as to highlight the target object regions, making the extracted feature more discriminative. Secondly, a transductive inference algorithm is proposed to alleviate the low-data problem, which iteratively utilizes the unlabeled query set to augment the support set, thereby making the class features more representative. Extensive experiments on two benchmarks show our method is a simple, effective and computationally efficient framework and outperforms the state-of-the-arts.
Image-text retrieval of natural scenes has been a popular research topic. Since image and text are heterogeneous cross-modal data, one of the key challenges is how to learn comprehensive yet unified representations to express the multi-modal data. A natural scene image mainly involves two kinds of visual concepts, objects and their relationships, which are equally essential to image-text retrieval. Therefore, a good representation should account for both of them. In the light of recent success of scene graph in many CV and NLP tasks for describing complex natural scenes, we propose to represent image and text with two kinds of scene graphs: visual scene graph (VSG) and textual scene graph (TSG), each of which is exploited to jointly characterize objects and relationships in the corresponding modality. The image-text retrieval task is then naturally formulated as cross-modal scene graph matching. Specifically, we design two particular scene graph encoders in our model for VSG and TSG, which can refine the representation of each node on the graph by aggregating neighborhood information. As a result, both object-level and relationship-level cross-modal features can be obtained, which favorably enables us to evaluate the similarity of image and text in the two levels in a more plausible way. We achieve state-of-the-art results on Flickr30k and MSCOCO, which verifies the advantages of our graph matching based approach for image-text retrieval.
Weakly supervised semantic segmentation has attracted much research interest in recent years considering its advantage of low labeling cost. Most of the advanced algorithms follow the design principle that expands and constrains the seed regions from class activation maps (CAM). As well-known, conventional CAM tends to be incomplete or over-activated due to weak supervision. Fortunately, we find that semantic segmentation has a characteristic of spatial transformation equivariance, which can form a few self-supervisions to help weakly supervised learning. This work mainly explores the advantages of scale equivariant constrains for CAM generation, formulated as a self-supervised scale equivariant network (SSENet). Specifically, a novel scale equivariant regularization is elaborately designed to ensure consistency of CAMs from the same input image with different resolutions. This novel scale equivariant regularization can guide the whole network to learn more accurate class activation. This regularized CAM can be embedded in most recent advanced weakly supervised semantic segmentation framework. Extensive experiments on PASCAL VOC 2012 datasets demonstrate that our method achieves the state-of-the-art performance both quantitatively and qualitatively for weakly supervised semantic segmentation. Code has been made available.
Visual Question Answering (VQA) is a challenging task for evaluating the ability of comprehensive understanding of the world. Existing benchmarks usually focus on the reasoning abilities either only on the vision or mainly on the knowledge with relatively simple abilities on vision. However, the ability of answering a question that requires alternatively inferring on the image content and the commonsense knowledge is crucial for an advanced VQA system. In this paper, we introduce a VQA dataset that provides more challenging and general questions about Compositional Reasoning on vIsion and Commonsense, which is named as CRIC. To create this dataset, we develop a powerful method to automatically generate compositional questions and rich annotations from both the scene graph of a given image and some external knowledge graph. Moreover, this paper presents a new compositional model that is capable of implementing various types of reasoning functions on the image content and the knowledge graph. Further, we analyze several baselines, state-of-the-art and our model on CRIC dataset. The experimental results show that the proposed task is challenging, where state-of-the-art obtains 52.26% accuracy and our model obtains 58.38%.