Gaze estimation, which seeks to reveal where a person is looking, provides a crucial clue for understanding human intentions and behaviors. Recently, Visual Transformer has achieved promising results in gaze estimation. However, dividing facial images into patches compromises the integrity of the image structure, which limits the inference performance. To tackle this challenge, we present Gaze-Swin, an end-to-end gaze estimation model formed with a dual-branch CNN-Transformer architecture. In Gaze-Swin, we adopt the Swin Transformer as the backbone network due to its effectiveness in handling long-range dependencies and extracting global features. Additionally, we incorporate a convolutional neural network as an auxiliary branch to capture local facial features and intricate texture details. To further enhance robustness and address overfitting issues in gaze estimation, we replace the original self-attention in the Transformer branch with Dropkey Assisted Attention (DA-Attention). In particular, this DA-Attention treats keys in the Transformer block as Dropout units and employs a decay Dropout rate schedule to preserve crucial gaze representations in deeper layers. Comprehensive experiments on three benchmark datasets demonstrate the superior performance of our method in comparison to the state of the art.
Recent research into video anomaly detection under weakly supervised settings has made significant progress in identifying anomalies with only coarse-grained annotations. Mainstream weakly supervised methods improve detection performance by generating high-quality pseudo labels for video segments. However, these pseudo-label-based methods have been ordinarily hindered by manually-set constraint rules as the bottleneck. In this paper, we propose the Feature Differentiation Reconstruction Network (FDR-Net), which no longer relies on pseudo labels and instead uses a differential reconstruction strategy to improve the discriminability of the representation. Concretely, video features are first randomly masked out and then reconstructed with distinct targets for normal and abnormal videos during the differential reconstruction process. Besides, we also introduce a dense transformer-based encoder to refine spatial-temporal relationships among video segments. Comprehensive experiments on ShanghaiTech demonstrate the superior performance of our model.
27 April 2024 in Journal of Visual Communication and Image Representation
Despite remarkable advancements, mainstream gaze estimation techniques, particularly appearance-based methods, often suffer from performance degradation in uncontrolled environments due to variations in illumination and individual facial attributes. Existing domain adaptation strategies, limited by their need for target domain samples, may fall short in realworld applications. This letter introduces Branch-out Auxiliary Regularization (BAR), an innovative method designed to boost gaze estimation's generalization capabilities without requiring direct access to target domain data. Specifically, BAR integrates two auxiliary consistency regularization branches: one that uses augmented samples to counteract environmental variations, and another that aligns gaze directions with positive source domain samples to encourage the learning of consistent gaze features. These auxiliary pathways strengthen the core network and are integrated into the original branch during training in a smooth, plug-and-play manner, facilitating easy adaptation to various other models without compromising the inference efficiency. Comprehensive experimental evaluations on four cross-dataset tasks demonstrate the superiority of our approach.
Recent years have witnessed significant progress in deep learning-based shadow removal. However, most prior methods operate primarily in the spatial domain or rely on coarse frequency cues, while the informative role of amplitude components in the frequency domain remains largely unexplored. The amplitude spectrum encodes spectral energy that reflects global illumination and fine texture that strongly influence shadow appearance. Motivated by this observation, we propose AmpFormer, a U-shaped transformer architecture that explicitly models amplitude information for robust shadow correction. Central to AmpFormer is a lightweight SFR module inserted at each encoder–decoder stage: SFR extracts multi-scale amplitude cues from compact spectral representations, learns per-channel adaptive gains and subtle phase adjustments, and injects the recalibrated frequency features into the spatial stream. To further encourage amplitude-aware restoration, we introduce an amplitude loss that explicitly regularizes spectral energy with emphasis on global illumination consistency. Extensive experiments on standard benchmarks demonstrate that AmpFormer achieves state-of-the-art restoration quality while offering a favorable computational-efficiency-accuracy trade-off, validating the practical benefit of amplitude-aware frequency modeling for shadow removal.