Video Classification on EV9V (test)
94.48AccuracyOurs
Evaluation results
| Method | Links | ||
|---|---|---|---|
| OursParams=14.75M, Input=5 × 3 × 224², FLOPs=17.04G2026.06 | 94.48 | 91.14 | |
| UniFormerV2 (B/16)Params=114.25M, Input=32 × 3 × 224², FLOPs=570.23G2026.06 | 94.26 | 90.9 | |
| TPN (R50)Params=91.50M, Input=80 × 3 × 256², FLOPs=547.50G2026.06 | 94.22 | 90.7 | |
| TANet (R50)Params=24.79M, Input=8 × 3 × 224², FLOPs=32.95G2026.06 | 94.07 | 90.45 | |
| TSN (R50)Params=23.53M, Input=8 × 3 × 224², FLOPs=32.87G2026.06 | 93.99 | 90.31 | |
| ResNet18 + LSTMParams=15.38M, Input=16 × 3 × 224², FLOPs=58.49G2026.06 | 93.92 | 90.54 | |
| MViT (Small)Params=34.24M, Input=80 × 3 × 224², FLOPs=322.28G2026.06 | 93.92 | 90.27 | |
| SlowOnly (R50)Params=31.65M, Input=80 × 3 × 256², FLOPs=547.50G2026.06 | 93.92 | 90.22 | |
| TSM (R50)Params=23.53M, Input=8 × 3 × 224², FLOPs=32.87G2026.06 | 93.66 | 89.95 | |
| ResNet18 + GRUParams=14.33M, Input=16 × 3 × 224², FLOPs=58.45G2026.06 | 93.54 | 90.1 | |
| Swin-T (P244)Params=27.86M, Input=128 × 3 × 224², FLOPs=352.25G2026.06 | 93.54 | 89.95 | |
| I3D (R50)Params=27.24M, Input=160 × 3 × 256², FLOPs=217.28G2026.06 | 93.32 | 89.26 | |
| R(2+1)D (R34)Params=63.76M, Input=80 × 3 × 256², FLOPs=531.41G2026.06 | 93.28 | 89.22 | |
| TIN (R50)Params=23.56M, Input=8 × 3 × 224², FLOPs=32.87G2026.06 | 93.21 | 89.16 | |
| ResNet18 + BiLSTMParams=21.69M, Input=16 × 3 × 224², FLOPs=58.69G2026.06 | 93.17 | 89.28 | |
| C2D (R50)Params=23.53M, Input=80 × 3 × 224², FLOPs=194.17G2026.06 | 92.91 | 88.72 | |
| C3DParams=78.03M, Input=160 × 3 × 112², FLOPs=385.47G2026.06 | 92.83 | 88.78 | |
| TimeSformerParams=121.27M, Input=8 × 3 × 224², FLOPs=196.05G2026.06 | 91.52 | 86.35 |