Video Classification on Sports-1M video level
70.8Clip Hit@1Conv Pooling
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Conv PoolingMethod=Image and Optical Flow, Frames=1202015.03 | 70.8 | 72.4 | 90.8 | — | — | |
| Convolution pooling on 120-frame clipsNumber of Nets=3 net2014.12 | 70.8 | — | — | 72.4 | 90.8 | |
| C3D (fine-tuned from I380K pre-trained model)Number of Nets=1 net, Training=fine-tuned from I380K pre-trained model2014.12 | 46.1 | — | — | 61.1 | 85.2 | |
| C3D (trained from scratch)Number of Nets=1 net, Training=from scratch2014.12 | 44.9 | — | — | 60 | 84.4 | |
| DeepVideo's Single-Frame + MultiresNumber of Nets=3 nets2014.12 | 42.4 | — | — | 60 | 78.5 | |
| Slow FusionCategory=Results [14], Frames=152015.03 | 41.9 | 60.9 | 80.2 | — | — | |
| DeepVideo's Slow FusionNumber of Nets=1 net2014.12 | 41.9 | — | — | 60.9 | 80.2 | |
| Single FrameCategory=Prior, Frames=12015.03 | 41.1 | 59.3 | 77.7 | — | — | |
| LSTMMethod=Image and Optical Flow, Frames=302015.03 | — | 73.1 | 90.5 | — | — |