Research map / Detection and segmentation / CNN features
Joint Semantic Instance Segmentation on Graphs with the Semantic Mutex Watershed
Steffen Wolf, Yuyan Li, Constantin Pape, Alberto Bailoni, Anna Kreshuk, Fred A. Hamprecht
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use joint partitioning and labeling (SMWS) instead of MWS followed by max semantic edge labeling | Mutex Watershed · 3D sponge EM volume (panoptic) | PQ: 48.1 -> 51.6 | Improved |
| use SMWS instead of connected components of semantic predictions | CCsem · 3D sponge EM volume (panoptic) | PQ: 43.4 -> 51.6 | Improved |
| use SMWS instead of connected components of short-range affinities | CCaff · 3D sponge EM volume (panoptic) | PQ: 24.3 -> 51.6 | Improved |
| use SMWS instead of the Asymmetric Multiway Cut (KLj*r) solver on largest feasible 2D slice | AMWC · Sponge 2D slice | PQ: 52.0 -> 49.2 | Hurt |
| use SMWS instead of PFPN with Mask-RCNN affinities and Deeplab semantics | PFPN · Cityscapes val (panoptic) | PQ: 58.1 -> 59.3 | Improved |
| use SMWS instead of SSAP | SSAP · Cityscapes val (panoptic) | PQ: 61.1 -> 59.3 | Hurt |
| use SMWS instead of AdaptIS | AdaptIS · Cityscapes val (panoptic) | PQ: 62.0 -> 59.3 | Hurt |
| use SMWS instead of UPSNet | UPSNet · Cityscapes val (panoptic) | PQ: 59.3 -> 59.3 | No clear effect |
| combine all affinity sources (Mask-RCNN, GMIS, Deeplab) instead of the best pair (ablation) | SMWS · Cityscapes val (panoptic) | PQ: 59.3 -> 46.3 | Hurt |
| derive affinities from Mask-RCNN foreground probabilities rather than other sources (ablation) | SMWS · Cityscapes val (panoptic) | PQ | Improved |
| replace AMWC solver by SMWS (p to infinity) for runtime | AMWC · Sponge 2D slices | runtime: 5 orders of magnitude faster | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 3 on most shared results.
- UPSNet — UPSNet: A Unified Panoptic Segmentation Network (CVPR 2019) · better on 1 of 3 shared results
- SSAP — SSAP: Single-Shot Instance Segmentation With Affinity Pyramid (ICCV 2019) · better on 0 of 2 shared results
- AdaptIS — AdaptIS: Adaptive Instance Selection Network (ICCV 2019) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- The Mutex Watershed and its Objective: Efficient, Parameter-Free Graph Partitioning (TPAMI 2020 · similarity 0.84)
- Instance-Aware Semantic Segmentation via Multi-Task Network Cascades (CVPR 2016 · similarity 0.84)
- InstanceCut: From Edges to Instances With MultiCut (CVPR 2017 · similarity 0.83)
- Kill Two Birds with One Stone: Domain Generalization for Semantic Segmentation via Network Pruning (IJCV 2024 · similarity 0.83)
- Structural Pruning via Spatial-aware Information Redundancy for Semantic Segmentation (AAAI 2025 · similarity 0.82)
- POT: Prototypical Optimal Transport for Weakly Supervised Semantic Segmentation (CVPR 2025 · similarity 0.82)
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.
