AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

SegmentMeIfYouCan: A Benchmark for Anomaly Segmentation

Robin Chan, Krzysztof Lis, Svenja Uhlemeyer, Hermann Blum, Sina Honari, Roland Siegwart, Pascal Fua, Mathieu Salzmann, Matthias Rottmann

NeurIPS 2021 · Detection and segmentation › Open-set and few-shot detection · approach: Interactive and referring segmentation · 69 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use a generative-model method rather than the classification-derived baselinesAnomaly and obstacle segmentation · RoadAnomaly21 and RoadObstacle21AuPRCImproved
Use maximized softmax entropy rather than image resynthesisAnomaly segmentation · RoadAnomaly21 testAuPRC: 52.3 -> 85.5Improved
Use maximized softmax entropy rather than SynBoostObstacle segmentation · RoadObstacle21 testAuPRC: 70.3 -> 85.1Improved
Use anomaly/obstacle proxy data to train the detectorAnomaly and obstacle segmentation · RoadAnomaly21 and RoadObstacle21AuPRCImproved
Apply feature-statistics methods under domain shiftMahalanobis and embedding density · RoadObstacle21AuPRCHurt
Use component-wise evaluation in addition to pixel-wise evaluationAnomaly segmentation evaluation · RoadAnomaly21 and RoadObstacle21component-wise average F1Mixed
Use camera-only anomaly methods on the LiDAR-guided small-obstacle datasetEvaluated anomaly segmentation methods · LiDAR guided Small Obstacle datasetperformanceHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy