andytang15/flag3dv2

FLAG3D++: A Benchmark for 3D Fitness Activity Comprehension with Language Instruction

The repository contains the official implementation of "FLAG3D++".

Yansong Tang¹   Aoyang Liu¹   Jinpeng Liu¹   Shiyi Zhang¹   Wenxun Dai¹   Jie Zhou¹   Xiu Li¹   Jiwen Lu^{1  
1}Tsinghua University

Set up

Environment

conda create -n hl-rac python==3.8

We recommend installing pytorch 1.14 to run the code.

pip install -r requirements.txt

Data

The whole data and corresponding weights can be obtained from this website. To make it easier to reproduce the results as well as to use them, you can directly use the already processed file (processed_data). If you have other scientific needs, you can also use raw files (raw_data).

Test

Please run following command:

cd hl-rac
python test.py

Acknowledgement

Our code is based on FLAG3D, 2s-AGCN, TransRAC, thanks to all the contributors!

flag3dv2's Issues

training code release?

hello, great work.
when will you release the training code?
thank you