dfk-paris / similarity

find similar/duplicate images

Requirements

• python >= 3.2
• ImageMagick >= 7

Installation

git clone https://github.com/dfk-paris/similARiTy
cd similARiTy
pip install -r requirements.txt

Usage

Basically, you first create an “index” for a given directory. This creates a json file. Do this for two directories and then compare the json files, e.g.:

bin/similar index /path/to/base_set base.json
bin/similar index /path/to/other_set other.json
bin/similar compare base.json other.json out.html

To list available sub commands, please run

bin/similar --help

and then show more specific help for each sub command with e.g.

bin/similar index --help

Hints

• Indexing is a lot faster if input images are small. A bounding box of 80x80 pixel is a good choice since it still allows to recognize the images in the html output.
• The default is the dhash with max distance 15. If you switch to phash, make sure to also reduce the distance, since the phash is only half the size of the dhash (64 bit vs 128 bit).
• phash is faster to compare since it is shorter than dhash.