Action Recognition on ActivityNet v1.3
96.9mAPText4Vis
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Text4VisBackbone=ViT-L, Input Resolution=3362022.07 | 96.9 | 93.3 | |
| Text4VisBackbone=ViT-L2022.07 | 96.5 | 92.9 | |
| NSNet2022.07 | 94.3 | 90.2 | |
| TSQNet2022.07 | 93.7 | 88.7 | |
| DSANet2022.07 | 90.5 | — | |
| DSA+I3DBackbone=3D ResNet50, Evaluation Protocol=M-TWI, Input Modality=RGB2021.05 | 90.5 | — | |
| MARLBackbone=SEResNeXt152, Pre-trained=Kinetics, Clips=252020.06 | 90.1 | 85.7 | |
| MARL2022.07 | 90.1 | 85.7 | |
| ListenToLook2022.07 | 89.9 | — | |
| V4D+I3DBackbone=3D ResNet50, Evaluation Protocol=M-TWI, Input Modality=RGB2021.05 | 88.9 | — | |
| R(2+1)D-RGB+FLOW & DSN (M = 1)Backbone=ResNet-34, Pre-trained=Kinetics, Clips=62020.06 | 87.8 | 82.1 | |
| R(2+1)D-RGB+FLOW & DSN (M = 1)Backbone=ResNet-34, Pre-trained=Kinetics, Clips=102020.06 | 87.8 | 82.6 | |
| R(2+1)D-RGB+FLOWBackbone=ResNet-34, Pre-trained=Kinetics, Clips=202020.06 | 87.1 | 81.3 | |
| TSN-Top3Backbone=Inception V3, Evaluation Protocol=M-TWI, Input Modality=RGB2021.05 | 84.5 | — | |
| R(2+1)D-FLOW & DSN (M = 1)Backbone=ResNet-34, Pre-trained=Kinetics, Clips=102020.06 | 84.3 | 79.1 | |
| MARLBackbone=ResNet-152, Pre-trained=ImageNet, Clips=252020.06 | 83.8 | 79.8 | |
| R(2+1)D-RGBBackbone=ResNet-34, Pre-trained=Kinetics, Clips=202020.06 | 83.6 | 78.2 | |
| MARLBackbone=ResNet-152, Pre-trained=Kinetics, Clips=252020.06 | 83.5 | 80.2 | |
| R(2+1)D-RGB & DSN (M = 1)Backbone=ResNet-34, Pre-trained=Kinetics, Clips=102020.06 | 83.5 | 78.5 | |
| RRABackbone=ResNet-152, Pre-trained=ImageNet, Clips=252020.06 | 83.4 | 78.8 | |
| TSNBackbone=Inception V3, Evaluation Protocol=M-TWI, Input Modality=RGB2021.05 | 83.3 | — | |
| R(2+1)D-FLOW & DSN (M = 1)Backbone=ResNet-34, Pre-trained=Kinetics, Clips=62020.06 | 82.9 | 78.2 | |
| R(2+1)D-RGB & DSN (M = 1)Backbone=ResNet-34, Pre-trained=Kinetics, Clips=62020.06 | 82.6 | 76.5 | |
| R(2+1)D-FLOWBackbone=ResNet-34, Pre-trained=Kinetics, Clips=202020.06 | 80.7 | 76.9 | |
| TSNBackbone=BN-Inception, Evaluation Protocol=M-TWI, Input Modality=RGB2021.05 | 79.7 | — | |
| P3DBackbone=ResNet-152, Pre-trained=ImageNet, Clips=202020.06 | 78.9 | 75.1 | |
| R3D-RGB & DSN (M = 6)Backbone=ResNet-18, Pre-trained=Kinetics, Clips=102020.06 | 71.7 | 68 | |
| R3D-RGB & DSN (M = 6)Backbone=ResNet-18, Pre-trained=Kinetics, Clips=62020.06 | 71.4 | 66.7 | |
| IDTPre-trained=ImageNet, Clips=ALL2020.06 | 68.7 | 64.7 | |
| R3D-RGB & TSN (6 seg)Backbone=ResNet-18, Pre-trained=Kinetics, Clips=202020.06 | 68.7 | 65.6 | |
| C3DPre-trained=Sports 1M, Clips=102020.06 | 67.7 | 65.8 | |
| Video-FocalNet-BPre-training=Kinetics4002023.07 | — | 89.8 | |
| Video-Swin-BPre-training=Kinetics4002023.07 | — | 88.5 |