iLearn-Lab/TOIS25-Awesome-Composed-Image-Retrieval

What it solves

This project tackles the challenge of keeping up with the quickly growing field of Composed Image Retrieval (CIR), where a search combines an image with textual feedback to find similar images. It gathers more than 120 papers from 2017 to 2025 and organizes them by technique, making it easier for researchers to find relevant work and understand the state of the art.

How it works

The repository is a manually curated list of papers and resources, classified into categories such as attribute-based CIR, supervised CIR, few-shot CIR, zero-shot CIR, and others. Each paper entry includes the publication venue, year, and a link to the paper. The project also offers a unified evaluation perspective by spanning roughly 10 benchmark datasets, and it encourages community contributions through pull requests and issues to keep the collection current.

Who it’s for

This is for researchers, students, and practitioners in computer vision, information retrieval, and multimodal learning. It serves as a convenient starting point to understand the CIR landscape, compare approaches, and identify open research problems.

Highlights

• Systematic analysis of 120+ seminal papers (2017–2025)
• Unified evaluation across ~10 benchmark datasets
• Continuously updated with community contributions (PRs and issues welcome)