This repository is a curated list of resources for practical metric learning and its applications, targeting researchers and engineers interested in developing more effective similarity-based systems. It provides a structured overview of surveys, applications, libraries, tools, and papers, aiming to inspire productivity and practical implementation in areas like natural language processing, computer vision, and recommendation systems.

How It Works

The list categorizes resources to cover the breadth of metric learning, from foundational surveys explaining mathematical concepts and traditional methods to cutting-edge deep learning approaches like contrastive learning and triplet loss. It highlights practical applications and libraries, showcasing how metric learning is used in real-world scenarios such as zero-shot classification, semantic search, and recommendation systems, along with tools for visualization and benchmarking.

Quick Start & Requirements

This is a curated list, not a runnable project. Links to specific libraries and demos are provided within the README for users to explore individual components.

Highlighted Details

• Comprehensive coverage of both traditional and deep metric learning techniques.
• Extensive examples of applications across NLP, CV, and RecSys domains.
• Links to key libraries like sentence-transformers, pytorch-metric-learning, and tensorflow-similarity.
• Inclusion of benchmarking tools and foundational papers like FaceNet and SimCLR.

Maintenance & Community

Maintained by Qdrant, with a stated goal to make metric learning more practical. Users are invited to join their Discord server for a paper reading club on metric learning. A contributing guide is available for those wishing to add to the list.

Licensing & Compatibility

The repository itself is a list and does not have a specific license. Individual libraries and applications linked within the list will have their own licenses, which users must consult.

Limitations & Caveats

This is an "awesome list" and not a unified framework or library. Users must individually assess and integrate the various resources and tools mentioned. The list is intended to be inspirational rather than exhaustive.