3D Single-Person Concurrent Activity Detection Using Stacked Relation Network

Authors: Yi Wei, Wenbo Li, Yanbo Fan, Linghan Xu, Ming-Ching Chang, Siwei Lyu12329-12337

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To evaluate the effectiveness of SRN, we conduct experiments on the UCLA concurrent activity dataset (Wei et al. 2013), which provides 3D human skeleton sequences with concurrent activity annotations.
Researcher Affiliation Collaboration 1University at Albany, State University of New York, USA 2Samsung Research America AI Center, USA 3Tencent AI Laboratory, China, 4Tianjin University, China
Pseudocode No The paper does not contain any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes The new dataset and code are available at https://github.com/weiyi1991/UA Concurrent/
Open Datasets Yes To evaluate the effectiveness of SRN, we conduct experiments on the UCLA concurrent activity dataset (Wei et al. 2013)... and The new dataset and code are available at https://github.com/weiyi1991/UA Concurrent/
Dataset Splits Yes We follow the same experimental setting as in (Wei et al. 2013) to use sequences with even indices for training and the remaining for test. We use two thirds of the sequences for training and the remaining for test.
Hardware Specification Yes All experiments are conducted on a machine with an NVIDIA GTX1080Ti GPU with 11GB onboard memory.
Software Dependencies No Our method is implemented with Py Torch. This statement mentions the software but lacks a specific version number, which is required for reproducibility.
Experiment Setup Yes minimized using back-propagation with ADAM optimizer and learning rate 0.001, betas (0.9, 0.999). We use Nr = 4 parallel relation modules within an Srelation layer. Dimension of the relation feature df is 64, and dk in (2) is set as 8. We set the hidden dimension for each LSTM binary classifier as 128.