Hi, Thanks for sharing the repo.
I want to train a model from scratch on the DISFA dataset. I have downloaded the dataset, which contains AU intensity values per frame in the AU_Labels folder. As part of data preprocessing, we need to feed these labels into dataset_process.py, which outputs two files:
- A file containing train/test/validation paths.
- A label file with values 0, 1, and -1, which is then used to generate the AU class weights.
My question is: Since the generated label file contains only 0, 1, and -1, does this mean we are using these binary values instead of the original AU intensity values (ranging from 0 to 5) for training the model?
Hi, Thanks for sharing the repo.
I want to train a model from scratch on the DISFA dataset. I have downloaded the dataset, which contains AU intensity values per frame in the AU_Labels folder. As part of data preprocessing, we need to feed these labels into
dataset_process.py, which outputs two files:My question is: Since the generated label file contains only 0, 1, and -1, does this mean we are using these binary values instead of the original AU intensity values (ranging from 0 to 5) for training the model?