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

Adaptive Image Transformations for Transfer-Based Adversarial Attack.- Generative Multiplane Images: Making a 2D GAN 3D-Aware.- AdvDO: Realistic Adversarial Attacks for Trajectory Prediction.- Adversarial Contrastive Learning via Asymmetric InfoNCE.- One Size Does NOT Fit All: Data-Adaptive Adversarial Training.- UniCR: Universally Approximated Certified Robustness via Randomized Smoothing.- Hardly Perceptible Trojan Attack against Neural Networks with Bit Flips.- Robust Network Architecture Search via Feature Distortion Restraining.- SecretGen: Privacy Recovery on Pre-trained Models via Distribution Discrimination.- Triangle Attack: A Query-Efficient Decision-Based Adversarial Attack.- Data-Free Backdoor Removal Based on Channel Lipschitzness.- Black-Box Dissector: Towards Erasing-Based Hard-Label Model Stealing Attack.- Learning Energy-Based Models with Adversarial Training.- Adversarial Label Poisoning Attack on Graph Neural Networks via Label Propagation.- Revisiting Outer Optimization in Adversarial Training.- Zero-Shot Attribute Attacks on Fine-Grained Recognition Models.- Towards Effective and Robust Neural Trojan Defenses via Input Filtering.- Scaling Adversarial Training to Large Perturbation Bounds.- Exploiting the Local Parabolic Landscapes of Adversarial Losses to Accelerate Black-Box Adversarial Attack.- Generative Domain Adaptation for Face Anti-Spoofing.- MetaGait: Learning to Learn an Omni Sample Adaptive Representation for Gait Recognition.- GaitEdge: Beyond Plain End-to-End Gait Recognition for Better Practicality.- UIA-ViT: Unsupervised Inconsistency-Aware Method Based on Vision Transformer for Face Forgery Detection.- Effective Presentation Attack Detection Driven by Face Related Task.- PPT: Token-Pruned Pose Transformer for Monocular and Multi-View Human Pose Estimation.- AvatarPoser: Articulated Full-Body Pose Tracking from Sparse Motion Sensing.- P-STMO: Pre-trained Spatial Temporal Many-to-One Model for 3D Human Pose Estimation.- D&D: Learning Human Dynamics from Dynamic Camera.- Explicit Occlusion Reasoning for Multi-Person 3D Human Pose Estimation.- COUCH: Towards Controllable Human-Chair Interactions.- Identity-Aware Hand Mesh Estimation and Personalization from RGB Images.- C3P: Cross-Domain Pose Prior Propagation for Weakly Supervised 3D Human Pose Estimation.- Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields.- CLIFF: Carrying Location Information in Full Frames into Human Pose and Shape Estimation.- DeciWatch: A Simple Baseline for 10Ö Efficient 2D and 3D Pose Estimation.- SmoothNet: A Plug-and-Play Network for Refining Human Poses in Videos.- PoseTrans: A Simple yet Effective Pose Transformation Augmentation for Human Pose Estimation.- Multi-Person 3D Pose and Shape Estimationvia Inverse Kinematics and Refinement.- Overlooked Poses Actually Make Sense: Distilling Privileged Knowledge for Human Motion Prediction.- Structural Triangulation: A Closed-Form Solution to Constrained 3D Human Pose Estimation.- Audio-Driven Stylized Gesture Generation with Flow-Based Model.- Self-Constrained Inference Optimization on Structural Groups for Human Pose Estimation