Anurag Bagchi, Jazib Mahmood, Dolton Fernandes, Ravi Kiran Sarvadevabhatla
State of the art architectures for untrimmed video Temporal Action Localization (TAL) have only considered RGB and Flow modalities, leaving the information-rich audio modality totally unexploited. Audio fusion has been explored for the related but arguably easier problem of trimmed (clip-level) action recognition. However, TAL poses a unique set of challenges. In this paper, we propose simple but effective fusion-based approaches for TAL. To the best of our knowledge, our work is the first to jointly consider audio and video modalities for supervised TAL. We experimentally show that our schemes consistently improve performance for state of the art video-only TAL approaches. Specifically, they help achieve new state of the art performance on large-scale benchmark datasets - ActivityNet-1.3 (54.34 mAP@0.5) and THUMOS14 (57.18 mAP@0.5). Our experiments include ablations involving multiple fusion schemes, modality combinations and TAL architectures. Our code, models and associated data are available at https://github.com/skelemoa/tal-hmo.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Video | ActivityNet-1.3 | mAP | 36.82 | AVFusion |
| Video | ActivityNet-1.3 | mAP IOU@0.5 | 54.34 | AVFusion |
| Video | ActivityNet-1.3 | mAP IOU@0.75 | 37.66 | AVFusion |
| Video | ActivityNet-1.3 | mAP IOU@0.95 | 8.93 | AVFusion |
| Video | THUMOS’14 | Avg mAP (0.3:0.7) | 53.3 | AVFusion |
| Video | THUMOS’14 | mAP IOU@0.3 | 70.1 | AVFusion |
| Video | THUMOS’14 | mAP IOU@0.4 | 64.9 | AVFusion |
| Video | THUMOS’14 | mAP IOU@0.5 | 57.1 | AVFusion |
| Video | THUMOS’14 | mAP IOU@0.6 | 45.4 | AVFusion |
| Video | THUMOS’14 | mAP IOU@0.7 | 28.8 | AVFusion |
| Video | THUMOS'14 | mAP IOU@0.5 | 57.18 | AVFusion |
| Temporal Action Localization | ActivityNet-1.3 | mAP | 36.82 | AVFusion |
| Temporal Action Localization | ActivityNet-1.3 | mAP IOU@0.5 | 54.34 | AVFusion |
| Temporal Action Localization | ActivityNet-1.3 | mAP IOU@0.75 | 37.66 | AVFusion |
| Temporal Action Localization | ActivityNet-1.3 | mAP IOU@0.95 | 8.93 | AVFusion |
| Temporal Action Localization | THUMOS’14 | Avg mAP (0.3:0.7) | 53.3 | AVFusion |
| Temporal Action Localization | THUMOS’14 | mAP IOU@0.3 | 70.1 | AVFusion |
| Temporal Action Localization | THUMOS’14 | mAP IOU@0.4 | 64.9 | AVFusion |
| Temporal Action Localization | THUMOS’14 | mAP IOU@0.5 | 57.1 | AVFusion |
| Temporal Action Localization | THUMOS’14 | mAP IOU@0.6 | 45.4 | AVFusion |
| Temporal Action Localization | THUMOS’14 | mAP IOU@0.7 | 28.8 | AVFusion |
| Temporal Action Localization | THUMOS'14 | mAP IOU@0.5 | 57.18 | AVFusion |
| Zero-Shot Learning | ActivityNet-1.3 | mAP | 36.82 | AVFusion |
| Zero-Shot Learning | ActivityNet-1.3 | mAP IOU@0.5 | 54.34 | AVFusion |
| Zero-Shot Learning | ActivityNet-1.3 | mAP IOU@0.75 | 37.66 | AVFusion |
| Zero-Shot Learning | ActivityNet-1.3 | mAP IOU@0.95 | 8.93 | AVFusion |
| Zero-Shot Learning | THUMOS’14 | Avg mAP (0.3:0.7) | 53.3 | AVFusion |
| Zero-Shot Learning | THUMOS’14 | mAP IOU@0.3 | 70.1 | AVFusion |
| Zero-Shot Learning | THUMOS’14 | mAP IOU@0.4 | 64.9 | AVFusion |
| Zero-Shot Learning | THUMOS’14 | mAP IOU@0.5 | 57.1 | AVFusion |
| Zero-Shot Learning | THUMOS’14 | mAP IOU@0.6 | 45.4 | AVFusion |
| Zero-Shot Learning | THUMOS’14 | mAP IOU@0.7 | 28.8 | AVFusion |
| Zero-Shot Learning | THUMOS'14 | mAP IOU@0.5 | 57.18 | AVFusion |
| Action Localization | ActivityNet-1.3 | mAP | 36.82 | AVFusion |
| Action Localization | ActivityNet-1.3 | mAP IOU@0.5 | 54.34 | AVFusion |
| Action Localization | ActivityNet-1.3 | mAP IOU@0.75 | 37.66 | AVFusion |
| Action Localization | ActivityNet-1.3 | mAP IOU@0.95 | 8.93 | AVFusion |
| Action Localization | THUMOS’14 | Avg mAP (0.3:0.7) | 53.3 | AVFusion |
| Action Localization | THUMOS’14 | mAP IOU@0.3 | 70.1 | AVFusion |
| Action Localization | THUMOS’14 | mAP IOU@0.4 | 64.9 | AVFusion |
| Action Localization | THUMOS’14 | mAP IOU@0.5 | 57.1 | AVFusion |
| Action Localization | THUMOS’14 | mAP IOU@0.6 | 45.4 | AVFusion |
| Action Localization | THUMOS’14 | mAP IOU@0.7 | 28.8 | AVFusion |
| Action Localization | THUMOS'14 | mAP IOU@0.5 | 57.18 | AVFusion |