5th ABAW Competition

1. Create environment

Create a python environment using conda, install packages in requirements.txt with pip install -r requirements.txt, or manually.

2. Preprocess data and extract feature

• extract wavs from videos using extract_wav.py.
• extract visual features, audio features using pretrained models.
• construct samples dataset using construct_*.py.
• split samples dataset to segments using split_*.py.

3. Train model

Train model with VA/solver.py, EXPR/solver.py and AU/solver.py.

4. Predict

Predict with VA/test.py, EXPR/test.py and AU/test.py.