Low-Resolution Action Recognition for Tiny Actions Challenge
This work addresses the practical challenge of activity recognition in real-world surveillance scenarios where videos are low-resolution and categories are imbalanced, representing an incremental improvement.
The paper tackles the problem of recognizing human activities in low-resolution surveillance videos with long-tailed category distributions by proposing a comprehensive solution including data-balanced training, dual-resolution distillation, and model ensemble with post-processing, achieving Top-1 ranking on the Tiny Actions Challenge leaderboard.
Tiny Actions Challenge focuses on understanding human activities in real-world surveillance. Basically, there are two main difficulties for activity recognition in this scenario. First, human activities are often recorded at a distance, and appear in a small resolution without much discriminative clue. Second, these activities are naturally distributed in a long-tailed way. It is hard to alleviate data bias for such heavy category imbalance. To tackle these problems, we propose a comprehensive recognition solution in this paper. First, we train video backbones with data balance, in order to alleviate overfitting in the challenge benchmark. Second, we design a dual-resolution distillation framework, which can effectively guide low-resolution action recognition by super-resolution knowledge. Finally, we apply model en-semble with post-processing, which can further boost per-formance on the long-tailed categories. Our solution ranks Top-1 on the leaderboard.