Skip to content

Latest commit

 

History

History
21 lines (15 loc) · 675 Bytes

README.md

File metadata and controls

21 lines (15 loc) · 675 Bytes
Raw

Video Recognition with UniRepLKNet

In this part, we explore utilizing a single convnet for video recognition task. Without special design, we just simply reshape the input tensors, and UniRepLKNet can deliver comparable performances better than other generlalist models. This part of code is based on VideoMAE. Thanks for their outstanding project.

Usage

1. Environment Setup.

pip install -r requirements.txt

2. Train and evaluate model.

We provide the experiment scripts for easier use:

  • Kinetics-400 Dataset
bash run_unireplknet.sh

Please edit run_unireplknet.sh before running the code.