MotivNet is a model built on Meta-Sapiens and ML-Decoder for the FER task on AffectNet.
We provide code to run inference on the model and code to train the model on your own data with custom specifications.
For inference, the output labels are shown below
0: Neutral, 1: Happiness, 2: Sadness, 3: Surprise, 4: Fear, 5: Disgust, 6: Anger, 7: Contempt.
Getting started is very simple.
First, create a new conda environment and run pip install -r requirements.txt
Then, download the MotivNet checkpoint from OneDrive. https://buckeyemailosu-my.sharepoint.com/:u:/g/personal/medicharla_2_buckeyemail_osu_edu/EfnsSxS42JNDipAEU45o-bUBsGfXviOOgaWka5LBLBkvBA?e=1SF6v8
Place this checkpoint in the /checkpoints/ folder to start finetuning or running inference on the model