Why Can't I Dance in the Mall? Learning to Mitigate Scene Bias in Action Recognition

Abstract

Human activities often occur in specific scene contexts, e.g., playing basketball on a basketball court. Training a model using existing video datasets thus inevitably captures and leverages such bias (instead of using the actual discriminative cues). The learned representation may not generalize well to new action classes or different tasks. In this paper, we propose to mitigate scene bias for video representation learning. Specifically, we augment the standard cross-entropy loss for action classification with 1) an adversarial loss for scene types and 2) a human mask confusion loss for videos where the human actors are masked out. These two losses encourage learning representations that are unable to predict the scene types and the correct actions when there is no evidence. We validate the effectiveness of our method by transferring our pre-trained model to three different tasks, including action classification, temporal localization, and spatio-temporal action detection. Our results show consistent improvement over the baseline model without debiasing.

NeurIPS 2019. Project webpage: http://chengao.vision/SDN/ Code: https://github.com/vt-vl-lab/SDN

Cited by in corpus (21)

• TCLR: Temporal Contrastive Learning for Video Representation
• A Comprehensive Study of Deep Video Action Recognition
• Learning from Failure: Training Debiased Classifier from Biased Classifier
• Dual Contrastive Learning for Spatio-temporal Representation
• MIST: Multiple Instance Self-Training Framework for Video Anomaly Detection
• Enhancing Unsupervised Video Representation Learning by Decoupling the Scene and the Motion
• Removing the Background by Adding the Background: Towards Background Robust Self-supervised Video Representation Learning
• Spatio-Temporal Perturbations for Video Attribution
• Evidential Deep Learning for Open Set Action Recognition
• Adaptive Label Smoothing
• Temporal Distinct Representation Learning for Action Recognition
• Learning Temporally Invariant and Localizable Features via Data Augmentation for Video Recognition
• Motion-aware Contrastive Video Representation Learning via Foreground-background Merging
• Metadata Normalization
• Hopper: Multi-hop Transformer for Spatiotemporal Reasoning
• Recent Progress in Appearance-based Action Recognition
• Gradient Frequency Modulation for Visually Explaining Video Understanding Models
• Learning Robust Convolutional Neural Networks with Relevant Feature Focusing via Explanations
• Nuisance-Label Supervision: Robustness Improvement by Free Labels
• PreViTS: Contrastive Pretraining with Video Tracking Supervision
• Depth Guided Adaptive Meta-Fusion Network for Few-shot Video Recognition