Papers › On the Unreasonable Effectiveness of Centroids in Image Retrieval

On the Unreasonable Effectiveness of Centroids in Image Retrieval

28 Apr 2021arXiv:2104.13643archive 2025-07-28

Mikolaj Wieczorek, Barbara Rychalska, Jacek Dabrowski

Image retrieval task consists of finding similar images to a query image from a set of gallery (database) images. Such systems are used in various applications e.g. person re-identification (ReID) or visual product search. Despite active development of retrieval models it still remains a challenging task mainly due to large intra-class variance caused by changes in view angle, lighting, background clutter or occlusion, while inter-class variance may be relatively low. A large portion of current research focuses on creating more robust features and modifying objective functions, usually based on Triplet Loss. Some works experiment with using centroid/proxy representation of a class to alleviate problems with computing speed and hard samples mining used with Triplet Loss. However, these approaches are used for training alone and discarded during the retrieval stage. In this paper we propose to use the mean centroid representation both during training and retrieval. Such an aggregated representation is more robust to outliers and assures more stable features. As each class is represented by a single embedding - the class centroid - both retrieval time and storage requirements are reduced significantly. Aggregating multiple embeddings results in a significant reduction of the search space due to lowering the number of candidate target vectors, which makes the method especially suitable for production deployments. Comprehensive experiments conducted on two ReID and Fashion Retrieval datasets demonstrate effectiveness of our method, which outperforms the current state-of-the-art. We propose centroid training and retrieval as a viable method for both Fashion Retrieval and ReID applications.

PaperPDFCode

Code

lannguyen0910/deep-efficient-reid mentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Image RetrievalPerson Re-IdentificationRetrieval

1 archive task tag without a task page not shown.

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Retrieval DeepFashion - Consumer-to-shop CTL Model (ResNet50-IBN-A, 320x320) Rank-1 37.3 #1 of 4 Archive leaderboard report
Image Retrieval DeepFashion - Consumer-to-shop CTL Model (ResNet50-IBN-A, 320x320) Rank-10 71.2 #1 of 4 Archive leaderboard report
Image Retrieval DeepFashion - Consumer-to-shop CTL Model (ResNet50-IBN-A, 320x320) Rank-20 77.7 #1 of 4 Archive leaderboard report
Image Retrieval DeepFashion - Consumer-to-shop CTL Model (ResNet50-IBN-A, 320x320) Rank-50 85.0 #1 of 4 Archive leaderboard report
Image Retrieval DeepFashion - Consumer-to-shop CTL Model (ResNet50-IBN-A, 320x320) mAP 49.2 #1 of 4 Archive leaderboard report
Image Retrieval DeepFashion - Consumer-to-shop CTL Model (ResNet50, 256x128) Rank-1 29.4 #3 of 4 Archive leaderboard report
Image Retrieval DeepFashion - Consumer-to-shop CTL Model (ResNet50, 256x128) Rank-10 61.3 #3 of 4 Archive leaderboard report
Image Retrieval DeepFashion - Consumer-to-shop CTL Model (ResNet50, 256x128) Rank-20 68.9 #3 of 4 Archive leaderboard report
Image Retrieval DeepFashion - Consumer-to-shop CTL Model (ResNet50, 256x128) Rank-50 77.4 #3 of 4 Archive leaderboard report
Image Retrieval DeepFashion - Consumer-to-shop CTL Model (ResNet50, 256x128) mAP 40.4 #3 of 4 Archive leaderboard report
Image Retrieval Exact Street2Shop CTL Model (ResNet50-IBN-A, 320x320) Rank-1 53.7 #1 of 3 Archive leaderboard report
Image Retrieval Exact Street2Shop CTL Model (ResNet50-IBN-A, 320x320) Rank-10 70.9 #1 of 3 Archive leaderboard report
Image Retrieval Exact Street2Shop CTL Model (ResNet50-IBN-A, 320x320) Rank-20 75.0 #1 of 3 Archive leaderboard report
Image Retrieval Exact Street2Shop CTL Model (ResNet50-IBN-A, 320x320) Rank-50 79.2 #1 of 3 Archive leaderboard report
Image Retrieval Exact Street2Shop CTL Model (ResNet50-IBN-A, 320x320) mAP 59.8 #1 of 3 Archive leaderboard report
Image Retrieval Exact Street2Shop CTL Model (ResNet50, 256x128) Rank-1 43.2 #3 of 3 Archive leaderboard report
Image Retrieval Exact Street2Shop CTL Model (ResNet50, 256x128) Rank-10 61.9 #3 of 3 Archive leaderboard report
Image Retrieval Exact Street2Shop CTL Model (ResNet50, 256x128) Rank-20 66.0 #3 of 3 Archive leaderboard report
Image Retrieval Exact Street2Shop CTL Model (ResNet50, 256x128) Rank-50 72.1 #3 of 3 Archive leaderboard report
Image Retrieval Exact Street2Shop CTL Model (ResNet50, 256x128) mAP 49.8 #3 of 3 Archive leaderboard report
Person Re-Identification DukeMTMC-reID CTL Model (ResNet50, 256x128) Rank-1 95.6 #2 of 94 Archive leaderboard report
Person Re-Identification DukeMTMC-reID CTL Model (ResNet50, 256x128) Rank-10 97.9 #2 of 94 Archive leaderboard report
Person Re-Identification DukeMTMC-reID CTL Model (ResNet50, 256x128) Rank-5 96.2 #2 of 94 Archive leaderboard report
Person Re-Identification DukeMTMC-reID CTL Model (ResNet50, 256x128) mAP 96.1 #2 of 94 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Triplet Loss

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections