Table 2 Performances on the optical flow of different CNNs
From: Improved two-stream model for human action recognition
Top 1% | Top 5% | |
|---|---|---|
ResNet101 [17] | 76.1 | |
VGG16 | 30.1 | 46.5 |
DenseNet121 (proposed method) | 79.63 | 80.12 |
From: Improved two-stream model for human action recognition
Top 1% | Top 5% | |
|---|---|---|
ResNet101 [17] | 76.1 | |
VGG16 | 30.1 | 46.5 |
DenseNet121 (proposed method) | 79.63 | 80.12 |