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

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

3 Aug 2020arXiv:2008.01232archive 2025-07-28

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.

Paper PDF Code

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

artest08/LateTemporalModeling3DCNN officialmentioned in papermentioned on GitHubpytorch report

kietngt00/hmdb51-recognition mentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Action Recognition

Results from the paper archive 2025-07-28

Task	Dataset	Model	Metric	Value	Rank at snapshot	Leaderboard	Report
Action Recognition	HMDB-51	R2+1D-BERT	Average accuracy of 3 splits	85.10	#7 of 77	Archive leaderboard	report
Action Recognition	UCF 101	R2+1D-BERT	3-fold Accuracy	98.69	#1 of 1	Archive leaderboard	report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

(2+1)D Convolution 1x1 Convolution 3D Convolution Average Pooling Batch Normalization Convolution Dense Connections Global Average Pooling Grouped Convolution Kaiming Initialization R(2+1)D ReLU ResNeXt ResNeXt Block Residual Connection

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections