Emotion Dataset Clip-Dissect

Using FER2013 and Affectnet image dataset trained model, applied clip-dissect to further understand the intermediate neurons and their roles.

To run the code, download prerequisites and installations. After running the installations, you will be able to run the ipynb files!

Hardware Set up

CPU: 8CPU

RAM: 16GB

GPU: 1GPU

Jupyterhub environment

Prerequisites

• Python 3.9 or higher: Download from Python's official website.

Installation

1. Clone the repository:

   git clone https://github.com/Trustworthy-ML-Lab/CLIP-dissect
   

Furthermore, you can install our conda environment with the environment.yml file.

cat environment.yml

conda env create --name new_env_name -f environment.yml

Dataset

FER2013

Image Download Link

Citations

• CLIP by OpenAI: GitHub Repository
• Original CLIP-dissect Project by the Trustworthy Machine Learning Lab: GitHub Repository