ARC Perceptual Hash

An investigatory repo exploring the relationship between perceptual hashing and in-painting

Installation

Dev

Ensure you have uv installed then run

uv sync --all-extras

Install the pre-commit hooks with

uv run pre-commit install

and run with

uv run pre-commit run --all-files

Usage

Dataset generation

See separate README.

Example usage (will download model if not run before):

uv run src/arc_phash/data_generation/perform_inpainting.py people runwayml/stable-diffusion-inpainting

Pipeline prototype

Ensure you have docker installed and running, then build the image with e.g.

docker build -t my_pipeline .

and run it with

docker run --rm --volume .:/app my_pipeline

where we have used --volume to create a bind mount which will save off a database named embeddings.db locally for querying and EDA. For example, run

sqlite3 embeddings.db

Then, to view a sample of rows:

SELECT * FROM embeddings LIMIT 10;