Multi-label Video Classification on Charades (val)
44.9mAPKSSNet
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| KSSNetBackbone=Inception-I3D, Modality=RGB, Pretrain=ImageNet2019.11 | 44.9 | 127 | |
| SlowFast(NL)Backbone=ResNet101, Modality=RGB, Pretrain=Kinetics-4002019.11 | 42.5 | 234 | |
| LFB(NL)Backbone=ResNet101-I3D, Modality=RGB, Pretrain=Kinetics-4002019.11 | 42.5 | — | |
| SlowFastBackbone=ResNet101, Modality=RGB, Pretrain=Kinetics-4002019.11 | 42.1 | 213 | |
| STRG (NL)Backbone=ResNet101-I3D, Modality=RGB, Pretrain=Kinetics-4002019.11 | 39.7 | 630 | |
| ResNet-101(NL)Backbone=ResNet101-I3D, Modality=RGB, Pretrain=Kinetics-4002019.11 | 37.5 | 544 | |
| I3DBackbone=Inception, Modality=RGB, Pretrain=Kinetics-4002019.11 | 32.9 | 108 | |
| MultiScale (TR)Backbone=Inception-I3D, Modality=RGB, Pretrain=ImageNet2019.11 | 25.2 | — | |
| CoViARModality=Compressed+Flow, Pretrain=ILSVRC2012-CLS2019.11 | 24.1 | — | |
| Asyn-TFBackbone=VGG16, Modality=RGB+Flow, Pretrain=ImageNet2019.11 | 22.4 | — | |
| CoViARModality=Compressed (Wu et al. 2018), Pretrain=ILSVRC2012-CLS2019.11 | 21.9 | — | |
| Two-streamBackbone=VGG16, Modality=RGB+Flow, Pretrain=ImageNet, UCF1012019.11 | 14.3 | — |