Exploring Stronger Feature for Temporal Action Localization
This work addresses feature quality limitations in video action localization for computer vision applications, but it is incremental as it combines existing methods.
The paper tackled the problem of temporal action localization by exploring stronger video features, finding that transformer-based backbones improve classification but not proposal accuracy, and achieved a 42.42% mAP on a validation set, which is 1.87% higher than a previous ensemble method.
Temporal action localization aims to localize starting and ending time with action category. Limited by GPU memory, mainstream methods pre-extract features for each video. Therefore, feature quality determines the upper bound of detection performance. In this technical report, we explored classic convolution-based backbones and the recent surge of transformer-based backbones. We found that the transformer-based methods can achieve better classification performance than convolution-based, but they cannot generate accuracy action proposals. In addition, extracting features with larger frame resolution to reduce the loss of spatial information can also effectively improve the performance of temporal action localization. Finally, we achieve 42.42% in terms of mAP on validation set with a single SlowFast feature by a simple combination: BMN+TCANet, which is 1.87% higher than the result of 2020's multi-model ensemble. Finally, we achieve Rank 1st on the CVPR2021 HACS supervised Temporal Action Localization Challenge.