Online pre-training with long-form videos
This work addresses action recognition for video analysis, but it appears incremental as it compares existing pre-training methods on a new data type.
The study tackled the problem of improving action recognition by pre-training models on long-form videos, finding that online pre-training with contrastive learning achieved the highest performance on downstream tasks.
In this study, we investigate the impact of online pre-training with continuous video clips. We will examine three methods for pre-training (masked image modeling, contrastive learning, and knowledge distillation), and assess the performance on downstream action recognition tasks. As a result, online pre-training with contrast learning showed the highest performance in downstream tasks. Our findings suggest that learning from long-form videos can be helpful for action recognition with short videos.