Papers › FlipReID: Closing the Gap between Training and Inference in Person Re-Identification

FlipReID: Closing the Gap between Training and Inference in Person Re-Identification

12 May 2021arXiv:2105.05639archive 2025-07-28

Xingyang Ni, Esa Rahtu

Since neural networks are data-hungry, incorporating data augmentation in training is a widely adopted technique that enlarges datasets and improves generalization. On the other hand, aggregating predictions of multiple augmented samples (i.e., test-time augmentation) could boost performance even further. In the context of person re-identification models, it is common practice to extract embeddings for both the original images and their horizontally flipped variants. The final representation is the mean of the aforementioned feature vectors. However, such scheme results in a gap between training and inference, i.e., the mean feature vectors calculated in inference are not part of the training pipeline. In this study, we devise the FlipReID structure with the flipping loss to address this issue. More specifically, models using the FlipReID structure are trained on the original images and the flipped images simultaneously, and incorporating the flipping loss minimizes the mean squared error between feature vectors of corresponding image pairs. Extensive experiments show that our method brings consistent improvements. In particular, we set a new record for MSMT17 which is the largest person re-identification dataset. The source code is available at https://github.com/nixingyang/FlipReID.

PaperPDFCode

Code

nixingyang/FlipReID officialmentioned in papermentioned on GitHubtf 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

Person Re-Identification

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Person Re-Identification DukeMTMC-reID FlipReID (with re-ranking) Rank-1 93.0 #9 of 94 Archive leaderboard report
Person Re-Identification DukeMTMC-reID FlipReID (with re-ranking) mAP 90.7 #9 of 94 Archive leaderboard report
Person Re-Identification DukeMTMC-reID FlipReID (without re-ranking) Rank-1 90.9 #36 of 94 Archive leaderboard report
Person Re-Identification DukeMTMC-reID FlipReID (without re-ranking) mAP 81.5 #36 of 94 Archive leaderboard report
Person Re-Identification MSMT17 FlipReID (with re-ranking) Rank-1 87.5 #3 of 43 Archive leaderboard report
Person Re-Identification MSMT17 FlipReID (with re-ranking) mAP 81.3 #3 of 43 Archive leaderboard report
Person Re-Identification MSMT17 FlipReID (without re-ranking) Rank-1 85.6 #17 of 43 Archive leaderboard report
Person Re-Identification MSMT17 FlipReID (without re-ranking) mAP 68.0 #17 of 43 Archive leaderboard report
Person Re-Identification Market-1501 FlipReID (with re-ranking) Rank-1 95.8 #38 of 135 Archive leaderboard report
Person Re-Identification Market-1501 FlipReID (with re-ranking) mAP 94.7 #38 of 135 Archive leaderboard report
Person Re-Identification Market-1501 FlipReID (without re-ranking) Rank-1 95.5 #54 of 135 Archive leaderboard report
Person Re-Identification Market-1501 FlipReID (without re-ranking) mAP 89.6 #54 of 135 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.

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