Semantic Neighborhood-Aware Deep Facial Expression Recognition

Cross-Layer Contrastive Learning of Latent Semantics for Facial Expression Recognition

Convolutional neural networks (CNNs) have achieved significant improvement for the task of facial expression recognition. However, current training still suffers from the inconsistent learning intensities among different layers, i.e., the feature representations in the shallow layers are not sufficiently learned compared with those in deep layers. To this end, this work proposes a contrastive learning framework to align the feature semantics of shallow and deep layers, followed by an attention module for representing the multi-scale features in the weight-adaptive manner. The proposed algorithm has three main merits. First, the learning intensity, defined as the magnitude of the backpropagation gradient, of the features on the shallow layer is enhanced by cross-layer contrastive learning. Second, the latent semantics in the shallow-layer and deep-layer features are explored and aligned in the contrastive learning, and thus the fine-grained characteristics of expressions can be taken into account for the feature representation learning. Third, by integrating the multi-scale features from multiple layers with an attention module, our algorithm achieved the state-of-the-art performances, i.e. 92.21%, 89.50%, 62.82%, on three in-the-wild expression databases, i.e. RAF-DB, FERPlus, SFEW, and the second best performance, i.e. 65.29% on AffectNet dataset. Our codes will be made publicly available.

Hierarchical attention network with progressive feature fusion for facial expression recognition

Facial expression recognition (FER) in the wild is challenging due to the disturbing factors including pose variation, occlusions, and illumination variation. The attention mechanism can relieve these issues by enhancing expression-relevant information and suppressing expression-irrelevant information. However, most methods utilize the same attention mechanism on feature tensors with varying spatial and channel sizes across different network layers, disregarding the dynamically changing sizes of these tensors. To solve this issue, this paper proposes a hierarchical attention network with progressive feature fusion for FER. Specifically, first, to aggregate diverse complementary features, a diverse feature extraction module based on several feature aggregation blocks is designed to exploit both local context and global context features, both low-level and high-level features, as well as the gradient features that are robust to illumination variation. Second, to effectively fuse the above diverse features, a hierarchical attention module (HAM) is designed to progressively enhance discriminative features from key parts of the facial images and suppress task-irrelevant features from disturbing facial regions. Extensive experiments show that our model achieves the best performance among existing FER methods.

Emotion Recognition from Large-Scale Video Clips with Cross-Attention and Hybrid Feature Weighting Neural Networks

The emotion of humans is an important indicator or reflection of their mental states, e.g., satisfaction or stress, and recognizing or detecting emotion from different media is essential to perform sequence analysis or for certain applications, e.g., mental health assessments, job stress level estimation, and tourist satisfaction assessments. Emotion recognition based on computer vision techniques, as an important method of detecting emotion from visual media (e.g., images or videos) of human behaviors with the use of plentiful emotional cues, has been extensively investigated because of its significant applications. However, most existing models neglect inter-feature interaction and use simple concatenation for feature fusion, failing to capture the crucial complementary gains between face and context information in video clips, which is significant in addressing the problems of emotion confusion and emotion misunderstanding. Accordingly, in this paper, to fully exploit the complementary information between face and context features, we present a novel cross-attention and hybrid feature weighting network to achieve accurate emotion recognition from large-scale video clips, and the proposed model consists of a dual-branch encoding (DBE) network, a hierarchical-attention encoding (HAE) network, and a deep fusion (DF) block. Specifically, the face and context encoding blocks in the DBE network generate the respective shallow features. After this, the HAE network uses the cross-attention (CA) block to investigate and capture the complementarity between facial expression features and their contexts via a cross-channel attention operation. The element recalibration (ER) block is introduced to revise the feature map of each channel by embedding global information. Moreover, the adaptive-attention (AA) block in the HAE network is developed to infer the optimal feature fusion weights and obtain the adaptive emotion features via a hybrid feature weighting operation. Finally, the DF block integrates these adaptive emotion features to predict an individual emotional state. Extensive experimental results of the CAER-S dataset demonstrate the effectiveness of our method, exhibiting its potential in the analysis of tourist reviews with video clips, estimation of job stress levels with visual emotional evidence, or assessments of mental healthiness with visual media.

Learning Deep Global Multi-Scale and Local Attention Features for Facial Expression Recognition in the Wild

Facial expression recognition (FER) in the wild received broad concerns in which occlusion and pose variation are two key issues. This paper proposed a global multi-scale and local attention network (MA-Net) for FER in the wild. Specifically, the proposed network consists of three main components: a feature pre-extractor, a multi-scale module, and a local attention module. The feature pre-extractor is utilized to pre-extract middle-level features, the multi-scale module to fuse features with different receptive fields, which reduces the susceptibility of deeper convolution towards occlusion and variant pose, while the local attention module can guide the network to focus on local salient features, which releases the interference of occlusion and non-frontal pose problems on FER in the wild. Extensive experiments demonstrate that the proposed MA-Net achieves the state-of-the-art results on several in-the-wild FER benchmarks: CAER-S, AffectNet-7, AffectNet-8, RAFDB, and SFEW with accuracies of 88.42%, 64.53%, 60.29%, 88.40%, and 59.40% respectively. The codes and training logs are publicly available at https://github.com/zengqunzhao/MA-Net.