Video Recognition on Something-Something V1
52.5AccuracyOursTDN
Evaluation results
| Method | Links | |
|---|---|---|
| OursTDNBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=312022.01 | 52.5 | |
| TDNBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=362022.01 | 52.3 | |
| STMBackbone=ResNet50, Pre-train=IN, #Frames=16x30, GFLOPS=67x302022.01 | 50.7 | |
| V4DBackbone=ResNet50, Pre-train=IN, #Frames=8x4, GFLOPS=167.62022.01 | 50.4 | |
| SmallBigNetBackbone=ResNet50, Pre-train=IN, #Frames=8+16, GFLOPS=1572022.01 | 50.4 | |
| TEINetBackbone=ResNet50, Pre-train=IN, #Frames=16, GFLOPS=662022.01 | 49.9 | |
| GSTBackbone=ResNet50, Pre-train=IN, #Frames=16, GFLOPS=592022.01 | 48.6 | |
| TAMBackbone=bLResNet50, Pre-train=IN, #Frames=16x2, GFLOPS=47.7x22022.01 | 48.4 | |
| S3D-GBackbone=Inception V1, Pre-train=IN, #Frames=64, GFLOPS=712022.01 | 48.2 | |
| AdaFocusBackbone=ResNet50, Pre-train=IN, #Frames=8+12, GFLOPS=33.72022.01 | 48.1 | |
| TANBackbone=ResNet50, Pre-train=IN, #Frames=16, GFLOPS=662022.01 | 47.6 | |
| TEINetBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=332022.01 | 47.4 | |
| GSTBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=292022.01 | 47 | |
| OursTSMBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=282022.01 | 47 | |
| AdaFuseBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=31.52022.01 | 46.8 | |
| TANBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=332022.01 | 46.5 | |
| ECO(En)Backbone=BNInception+3D ResNet18, Pre-train=K400, #Frames=92, GFLOPS=2672022.01 | 46.4 | |
| STMBackbone=ResNet50, Pre-train=IN, #Frames=8x30, GFLOPS=33x302022.01 | 46.4 | |
| RubiksNetBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=15.82022.01 | 46.4 | |
| I3D+GCN+NLBackbone=3D ResNet50, Pre-train=K400, #Frames=32x2, GFLOPS=303x22022.01 | 46.1 | |
| TSMBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=332022.01 | 45.6 | |
| CoConvBackbone=GoogleNet, Pre-train=IN, #Frames=16, GFLOPS=182022.01 | 45.5 | |
| Non-local I3DBackbone=3D ResNet50, Pre-train=K400, #Frames=32x2, GFLOPS=168x22022.01 | 44.4 | |
| I3DBackbone=3D ResNet50, Pre-train=K400, #Frames=32x2, GFLOPS=153x22022.01 | 41.6 | |
| TRNBackbone=BN-Inception, Pre-train=IN, #Frames=8, GFLOPS=162022.01 | 34.4 | |
| TSNBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=33.22022.01 | 19.7 | |
| AR-NetBackbone=ResNet50, Pre-train=IN, #Frames=8, GFLOPS=41.42022.01 | 18.9 |