Late Temporal Modeling in 3D CNN Architectures with BERT for Action Recognition

TL;DR

This paper enhances 3D CNN architectures for action recognition by integrating BERT for late temporal modeling, significantly improving accuracy on HMDB51 and UCF101 datasets.

Contribution

It introduces replacing TGAP with BERT in 3D CNNs, achieving state-of-the-art results in action recognition.

Findings

Improved accuracy on HMDB51 and UCF101 datasets.

BERT-based temporal modeling outperforms traditional pooling.

Applicable to multiple 3D CNN architectures.

Abstract

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.