ActionFormer (InternVideo features)

Reported on 4 benchmarks across 4 tasks · 1 paper · 4 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision3 results

VideoonTHUMOS’14
Avg mAP (0.3:0.7)· uses extra data· 2022-12-06
71.58
best: 76.9 (AdaTAD (VideoMAEv2-giant))
SOTA
InternVideo: General Video Foundation Models via Generative and Discriminative Learning arXiv:2212.03191
Temporal Action LocalizationonTHUMOS’14
Avg mAP (0.3:0.7)· uses extra data· 2022-12-06
71.58
best: 76.9 (AdaTAD (VideoMAEv2-giant))
SOTA
InternVideo: General Video Foundation Models via Generative and Discriminative Learning arXiv:2212.03191
Action LocalizationonTHUMOS’14
Avg mAP (0.3:0.7)· uses extra data· 2022-12-06
71.58
best: 76.9 (AdaTAD (VideoMAEv2-giant))
SOTA
InternVideo: General Video Foundation Models via Generative and Discriminative Learning arXiv:2212.03191

Methodology1 result

Zero-Shot LearningonTHUMOS’14
Avg mAP (0.3:0.7)· uses extra data· 2022-12-06
71.58
best: 76.9 (AdaTAD (VideoMAEv2-giant))
SOTA
InternVideo: General Video Foundation Models via Generative and Discriminative Learning arXiv:2212.03191