Facebook AI Image Similarity Challenge: Descriptor Track

This repository contains the code for our solution to the Facebook AI Image Similarity Challenge: Descriptor Track, hosted on DrivenData:
https://www.drivendata.org/competitions/85/competition-image-similarity-2-final/leaderboard/

The solution is described in the paper hosted on arXiv.org:
https://arxiv.org/abs/2112.03415v2

Environment Setup

To run the models, you must first configure the environment:

Download the datasets following the README.md instructions at: all_datasets/README.md
Copy some query images. Go to the directory FB_page_qry/ and run: python3 copy_fb_query_images.py

Download all the checkpoints from my gdrive and locate them inside checkpoints/ directory.

Follow the README.md instructions at: phase1_scripts/README.md

Follow the README.md instructions at: phase2_scripts/README.md

Follow the README.md instructions at: ensemble_training_scripts/README.md

Follow the README.md instructions at: training_scripts/README.md

Name		Name	Last commit message	Last commit date
Latest commit History 10 Commits
FB_page_qry		FB_page_qry
all_datasets		all_datasets
checkpoints		checkpoints
data		data
ensemble_training_scripts		ensemble_training_scripts
modules		modules
phase1_scripts		phase1_scripts
phase2_scripts		phase2_scripts
training_scripts		training_scripts
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
requirements.txt		requirements.txt