Beyond Pixel Overlap: A Framework for Decomposing Segmentation Evaluation Metrics
Not on the map. Why?
Abstract (arXiv)
Evaluation metrics are central to binary target segmentation because they determine how progress is measured, compared, and interpreted. In this paper, target denotes the task-defined positive region to be segmented rather than a generic foreground object. It may be salient, camouflaged, transparent, glass-like, mirror-like, shadow-like, lesion-like, or defined by other application-specific semantics. We treat existing metrics as compositions of modular design choices rather than isolated formulas. The proposed framework decomposes each metric into five stages covering prediction representation, target extraction, target matching, score computation, and metric reporting. We use this framework to analyze representative metrics and show how newer metrics address specific limits in earlier protocols. The stage choices keep each metric's assumptions visible. We then discuss the design space opened by the framework and its implications for task-aware evaluation protocols. Reference code is available at https://github.com/lartpang/PySODMetrics.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 30 of its 41 references that are in the map. First posted on arXiv 1 Jul 2026.
Builds on
30 of 41 references.
References · 30
- 2014Microsoft COCO: Common Objects in Contextcited by 17,608
- 2017Structure-measure: A New Way to Evaluate Foreground Mapscited by 480
- 2020PraNet: Parallel Reverse Attention Network for Polyp Segmentationcited by 460
- 2018Enhanced-alignment Measure for Binary Foreground Map Evaluationcited by 370
- 2021Anabranch Network for Camouflaged Object Segmentationcited by 227
- 2019A Simple Pooling-Based Design for Real-Time Salient Object Detectioncited by 212
- 2020Multi-scale Interactive Network for Salient Object Detectioncited by 212
- 2021Concealed Object Detectioncited by 212
- 2021Camouflaged Object Segmentation with Distraction Miningcited by 151
- 2022Zoom In and Out: A Mixed-scale Triplet Network for Camouflaged Object Detectioncited by 145
- 2020Suppress and Balance: A Simple Gated Network for Salient Object Detectioncited by 126
- 2024Bilateral Reference for High-Resolution Dichotomous Image Segmentationcited by 112
18 more
- 2022Highly Accurate Dichotomous Image Segmentationcited by 110
- 2014Salient Object Detection: A Surveycited by 98
- 2020Segmenting Transparent Objects in the Wildcited by 98
- 2021Automatic Polyp Segmentation via Multi-scale Subtraction Networkcited by 77
- 2020RGB-D Salient Object Detection: A Surveycited by 73
- 2020Hierarchical Dynamic Filtering Network for RGB-D Salient Object Detectioncited by 72
- 2020A Single Stream Network for Robust and Real-time RGB-D Salient Object Detectioncited by 67
- 2023ZoomNeXt: A Unified Collaborative Pyramid Network for Camouflaged Object Detectioncited by 48
- 2024A Survey of Camouflaged Object Detection and Beyondcited by 37
- 2024Multi-view Aggregation Network for Dichotomous Image Segmentationcited by 28
- 2024Spider: A Unified Framework for Context-dependent Concept Segmentationcited by 18
- 2024Size-invariance Matters: Rethinking Metrics and Losses for Imbalanced Multi-object Salient Object Detectioncited by 14
- 2023M2SNet: Multi-scale in Multi-scale Subtraction Network for Medical Image Segmentationcited by 13
- 2024Unveiling Deep Shadows: A Survey and Benchmark on Image and Video Shadow Detection, Removal, and Generation in the Deep Learning Eracited by 7
- 2025S3OD: Towards Generalizable Salient Object Detection with Synthetic Datacited by 6
- 2021Multiscale IoU: A Metric for Evaluation of Salient Object Detection with Fine Structurescited by 5
- 2025Rethinking Evaluation of Infrared Small Target Detectioncited by 4
- 2025Context-measure: Contextualizing Metric for Camouflagecited by 2
Built on by
Citing papers per year: 2026: 1.
1 paper.