Late Temporal Modeling in 3D CNN Architectures with BERT for Action Recognition
This work addresses action recognition for video analysis, offering incremental improvements by integrating BERT into existing 3D CNN architectures.
The paper tackles the problem of action recognition by combining 3D convolution with late temporal modeling, replacing Temporal Global Average Pooling with BERT to better utilize temporal information, resulting in state-of-the-art accuracies of 85.10% on HMDB51 and 98.69% on UCF101.
In this work, we combine 3D convolution with late temporal modeling for action recognition. For this aim, we replace the conventional Temporal Global Average Pooling (TGAP) layer at the end of 3D convolutional architecture with the Bidirectional Encoder Representations from Transformers (BERT) layer in order to better utilize the temporal information with BERT's attention mechanism. We show that this replacement improves the performances of many popular 3D convolution architectures for action recognition, including ResNeXt, I3D, SlowFast and R(2+1)D. Moreover, we provide the-state-of-the-art results on both HMDB51 and UCF101 datasets with 85.10% and 98.69% top-1 accuracy, respectively. The code is publicly available.