Papers › Unsupervised Person Re-identification: Clustering and Fine-tuning

Unsupervised Person Re-identification: Clustering and Fine-tuning

30 May 2017arXiv:1705.10444archive 2025-07-28

Hehe Fan, Liang Zheng, Yi Yang

The superiority of deeply learned pedestrian representations has been reported in very recent literature of person re-identification (re-ID). In this paper, we consider the more pragmatic issue of learning a deep feature with no or only a few labels. We propose a progressive unsupervised learning (PUL) method to transfer pretrained deep representations to unseen domains. Our method is easy to implement and can be viewed as an effective baseline for unsupervised re-ID feature learning. Specifically, PUL iterates between 1) pedestrian clustering and 2) fine-tuning of the convolutional neural network (CNN) to improve the original model trained on the irrelevant labeled dataset. Since the clustering results can be very noisy, we add a selection operation between the clustering and fine-tuning. At the beginning when the model is weak, CNN is fine-tuned on a small amount of reliable examples which locate near to cluster centroids in the feature space. As the model becomes stronger in subsequent iterations, more images are being adaptively selected as CNN training samples. Progressively, pedestrian clustering and the CNN model are improved simultaneously until algorithm convergence. This process is naturally formulated as self-paced learning. We then point out promising directions that may lead to further improvement. Extensive experiments on three large-scale re-ID datasets demonstrate that PUL outputs discriminative features that improve the re-ID accuracy.

PaperPDFCode

Code

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

ClusteringPerson Re-IdentificationUnsupervised Domain AdaptationUnsupervised Person Re-Identification

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Person Re-Identification DukeMTMC-reID PUL* Rank-1 30.4 #93 of 94 Archive leaderboard report
Person Re-Identification DukeMTMC-reID PUL* mAP 16.4 #93 of 94 Archive leaderboard report
Person Re-Identification Market-1501 PUL* Rank-1 44.7 #123 of 135 Archive leaderboard report
Person Re-Identification Market-1501 PUL* mAP 20.1 #123 of 135 Archive leaderboard report
Unsupervised Domain Adaptation VehicleID to VeRi-776 PUL Rank-1 55.24 #13 of 14 Archive leaderboard report
Unsupervised Domain Adaptation VehicleID to VeRi-776 PUL Rank-10 - #13 of 14 Archive leaderboard report
Unsupervised Domain Adaptation VehicleID to VeRi-776 PUL Rank-5 67.34 #13 of 14 Archive leaderboard report
Unsupervised Domain Adaptation VehicleID to VeRi-776 PUL mAP 17.06 #13 of 14 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Large PUL R-1 30.90 #10 of 13 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Large PUL R-10 - #10 of 13 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Large PUL R-5 47.18 #10 of 13 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Large PUL mAP 34.71 #10 of 13 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Medium PUL R-1 33.83 #10 of 13 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Medium PUL R-10 - #10 of 13 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Medium PUL R-5 49.72 #10 of 13 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Medium PUL mAP 37.68 #10 of 13 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Small PUL mAP 43.90 #6 of 8 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Small PUL R-1 40.03 #6 of 8 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Small PUL R-10 - #6 of 8 Archive leaderboard report
Unsupervised Domain Adaptation Veri-776 to VehicleID Small PUL R-5 56.03 #6 of 8 Archive leaderboard report
Unsupervised Person Re-Identification DukeMTMC-reID PUL MAP 16.4 #12 of 13 Archive leaderboard report
Unsupervised Person Re-Identification DukeMTMC-reID PUL Rank-1 30.0 #12 of 13 Archive leaderboard report
Unsupervised Person Re-Identification DukeMTMC-reID PUL Rank-10 48.5 #12 of 13 Archive leaderboard report
Unsupervised Person Re-Identification DukeMTMC-reID PUL Rank-5 43.4 #12 of 13 Archive leaderboard report
Unsupervised Person Re-Identification Market-1501 PUL MAP 20.5 #22 of 23 Archive leaderboard report
Unsupervised Person Re-Identification Market-1501 PUL Rank-1 45.5 #22 of 23 Archive leaderboard report
Unsupervised Person Re-Identification Market-1501 PUL Rank-10 66.7 #22 of 23 Archive leaderboard report
Unsupervised Person Re-Identification Market-1501 PUL Rank-5 60.7 #22 of 23 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