Computer Vision – ECCV 2022 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part IV

Action Recognition.- Dual-Evidential Learning for Weakly-Supervised Temporal Action

Localization.- Global-Local Motion Transformer for Unsupervised Skeleton-Based

Action Learning.- AdaFocusV3: On Unified Spatial-Temporal Dynamic Video Recognition.- Panoramic Human Activity Recognition.- Delving into Details: Synopsis-to-Detail Networks for Video Recognition.- A Generalized & Robust Framework for Timestamp Supervision in Temporal Action Segmentation.- Few-Shot Action Recognition with Hierarchical Matching and Contrastive Learning.- PrivHAR: Recognizing Human Actions from Privacy-Preserving Lens.- Scale-Aware Spatio-Temporal Relation Learning for Video Anomaly Detection.- Compound Prototype Matching for Few-Shot Action Recognition.- Continual 3D Convolutional Neural Networks for Real-Time Processing of Videos.- Dynamic Spatio-Temporal Specialization Learning for Fine-Grained Action Recognition.- Dynamic Local Aggregation Network with Adaptive Clusterer for Anomaly Detection.- Action Quality Assessment with Temporal Parsing Transformer.- Entry-Flipped Transformer for Inference and Prediction of Participant Behavior.- Pairwise Contrastive Learning Network for Action Quality Assessment.- Geometric Features Informed Multi-Person Human-Object Interaction Recognition in Videos.- ActionFormer: Localizing Moments of Actions with Transformers.- SocialVAE: Human Trajectory Prediction Using Timewise Latents.- Shape Matters: Deformable Patch Attack.- Frequency Domain Model Augmentation for Adversarial Attack.- Prior-Guided Adversarial Initialization for Fast Adversarial Training.- Enhanced Accuracy and Robustness via Multi-Teacher Adversarial Distillation.- LGV: Boosting Adversarial Example Transferability from Large Geometric Vicinity.- A Large-Scale Multiple-Objective Method for Black-Box Attack against Object Detection.- GradAuto: Energy-Oriented Attack on Dynamic Neural Networks.- A Spectral View of Randomized Smoothing under Common Corruptions: Benchmarking and Improving Certified Robustness.- Improving Adversarial Robustness of 3D Point Cloud Classification Models.- Learning Extremely Lightweight and Robust Model with Differentiable Constraints on Sparsity and Condition Number.- RIBAC: Towards Robust and Imperceptible Backdoor Attack against Compact DNN.- Boosting Transferability of Targeted Adversarial Examples via Hierarchical Generative Networks.