CIRCO

Composed Image Retrieval on Common Objects in context

ImagesTextsCreative Commons BY-NC 4.0Introduced 2023-03-27

CIRCO (Composed Image Retrieval on Common Objects in context) is an open-domain benchmarking dataset for Composed Image Retrieval (CIR) based on real-world images from COCO 2017 unlabeled set. It is the first CIR dataset with multiple ground truths and aims to address the problem of false negatives in existing datasets. CIRCO comprises a total of 1020 queries, randomly divided into 220 and 800 for the validation and test set, respectively, with an average of 4.53 ground truths per query.

Source: Zero-Shot Composed Image Retrieval with Textual Inversion

Image Source: Zero-Shot Composed Image Retrieval with Textual Inversion

Benchmarks

Composed Image Retrieval (CoIR)/mAP@10 Composed Image Retrieval (CoIR)/MAP@5 Composed Image Retrieval (CoIR)/mAP@50 Composed Image Retrieval (CoIR)/mAP@25 Image Retrieval/mAP@10 Image Retrieval/MAP@5 Image Retrieval/mAP@50 Image Retrieval/mAP@25

Related Benchmarks

CirCor DigiScope/Phonocardiogram Classification/Unweighted average recall CirCor DigiScope/Phonocardiogram Classification/Weighted Accuracy CirCor DigiScope/Phonocardiogram Classification/Weighted accuracy (cross-val)CirCor DigiScope/Phonocardiogram Classification/Weighted accuracy (validation)CirCor DigiScope/Predict clinical outcome/Clinical cost score CirCor DigiScope/Predict clinical outcome/Clinical cost score (cross-val)CirCor DigiScope/Predict clinical outcome/Clinical cost score (validation data)