TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK	EXTRA DATA	REMOVE
Action Recognition	HMDB-51	R2+1D-BERT	Average accuracy of 3 splits	85.10	# 6
Action Recognition	UCF 101	R2+1D-BERT	3-fold Accuracy	98.69	# 1

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/late-temporal-modeling-in-3d-cnn/action-recognition-on-ucf-101)](https://paperswithcode.com/sota/action-recognition-on-ucf-101?p=late-temporal-modeling-in-3d-cnn)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/late-temporal-modeling-in-3d-cnn/action-recognition-in-videos-on-hmdb-51)](https://paperswithcode.com/sota/action-recognition-in-videos-on-hmdb-51?p=late-temporal-modeling-in-3d-cnn)`

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

3 Aug 2020 · M. Esat Kalfaoglu, Sinan Kalkan, A. Aydin Alatan ·

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.

PDF Abstract