Papers › DSR -- A dual subspace re-projection network for surface anomaly detection

DSR -- A dual subspace re-projection network for surface anomaly detection

2 Aug 2022arXiv:2208.01521archive 2025-07-28

Vitjan Zavrtanik, Matej Kristan, Danijel Skočaj

The state-of-the-art in discriminative unsupervised surface anomaly detection relies on external datasets for synthesizing anomaly-augmented training images. Such approaches are prone to failure on near-in-distribution anomalies since these are difficult to be synthesized realistically due to their similarity to anomaly-free regions. We propose an architecture based on quantized feature space representation with dual decoders, DSR, that avoids the image-level anomaly synthesis requirement. Without making any assumptions about the visual properties of anomalies, DSR generates the anomalies at the feature level by sampling the learned quantized feature space, which allows a controlled generation of near-in-distribution anomalies. DSR achieves state-of-the-art results on the KSDD2 and MVTec anomaly detection datasets. The experiments on the challenging real-world KSDD2 dataset show that DSR significantly outperforms other unsupervised surface anomaly detection methods, improving the previous top-performing methods by 10% AP in anomaly detection and 35% AP in anomaly localization.

Paper PDF Code

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

vitjanz/dsr_anomaly_detection officialmentioned in paperpytorch 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

Anomaly Detection Anomaly Localization Defect Detection Supervised Defect Detection Unsupervised Anomaly Detection Weakly Supervised Defect Detection

Results from the paper archive 2025-07-28

Task	Dataset	Model	Metric	Value	Rank at snapshot	Leaderboard	Report
Anomaly Detection	MVTec AD	DSR	Detection AUROC	98.2	#65 of 148	Archive leaderboard	report
Anomaly Detection	MVTec AD	DSR	Segmentation AP	70.2	#65 of 148	Archive leaderboard	report
Anomaly Detection	MVTec LOCO AD	DSR	Avg. Detection AUROC	82.6	#24 of 40	Archive leaderboard	report
Anomaly Detection	MVTec LOCO AD	DSR	Detection AUROC (only logical)	75.0	#24 of 40	Archive leaderboard	report
Anomaly Detection	MVTec LOCO AD	DSR	Detection AUROC (only structural)	90.2	#24 of 40	Archive leaderboard	report
Anomaly Detection	MVTec LOCO AD	DSR	Segmentation AU-sPRO (until FPR 5%)	58.5	#24 of 40	Archive leaderboard	report
Anomaly Detection	VisA	DSR	Segmentation AUPRO (until 30% FPR)	68.1	#46 of 50	Archive leaderboard	report
Unsupervised Anomaly Detection	KolektorSDD2	DSR	Detection AP	87.2	#2 of 3	Archive leaderboard	report
Unsupervised Anomaly Detection	KolektorSDD2	DSR	Segmentation AP	61.4	#2 of 3	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