This is a reimplementation of the paper Similarity Search for Efficient Active Learning and Search of Rare Concepts The goal is to recreate Figure 1b in the paper using MaxEnt-SEALS.
I recommend setting up a new conda environment. Install the package requirements:
conda install -c pytorch faiss-gpu pip
pip install -r requirements.txt
python main.py
To run the manual labeling dashboard, run
python main_manual_labeling.py
You will see a link to a dashboard in the terminal which can be used for manual labeling.
Install the pre-commit hooks for linting:
pre-commit install