Research map / Detection and segmentation / Open-set and few-shot detection
The Mutex Watershed: Efficient, Parameter-Free Image Partitioning
Steffen Wolf, Constantin Pape, Alberto Bailoni, Nasim Rahaman, Anna Kreshuk, Ullrich Kothe, FredA. 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 Mutex Watershed instead of lifted multicut post-processing on the top leaderboard network | UNet · ISBI 2012 test leaderboard | Rand-Score: 0.98788 -> 0.98792 | Improved |
| use Mutex Watershed instead of MC-FULL multicut on same CNN | CNN affinities · ISBI 2012 (same CNN predictions, post-processing comparison) | Rand-Score: 0.98029 -> 0.98792 | Improved |
| use Mutex Watershed instead of lifted multicut on same CNN | CNN affinities · ISBI 2012 (same CNN predictions, post-processing comparison) | Rand-Score: 0.97990 -> 0.98792 | Improved |
| use Mutex Watershed instead of MC-FULL for runtime | CNN affinities · ISBI 2012 (same CNN predictions, post-processing comparison) | runtime (s): 9415.8 -> 43.32 | Improved |
| add long-range repulsive predictions (MWS/MC-FULL/LMC versus short-range only MC-LOCAL) (ablation) | multicut · ISBI 2012 (same CNN predictions, post-processing comparison) | Rand-Score: 0.70990 -> 0.98029 | Improved |
| use Mutex Watershed instead of plain watershed | CNN affinities · ISBI 2012 (same CNN predictions, post-processing comparison) | Rand-Score: 0.63958 -> 0.98792 | Improved |
| use Mutex Watershed instead of thresholding | CNN affinities · ISBI 2012 (same CNN predictions, post-processing comparison) | Rand-Score: 0.91435 -> 0.98792 | Improved |
| replace network affinities by ground-truth affinities (ablation) | MWS · BSD500 test (MWS on interpolated affinities) | Rand index: 0.826 -> 0.901 | Improved |
| add 62% noise to ground-truth affinities (ablation) | MWS · BSD500 test (MWS on interpolated affinities) | Rand index: 0.901 -> 0.897 | No clear effect |
| add 66% noise to ground-truth affinities (ablation) | MWS · BSD500 test (MWS on interpolated affinities) | Rand index: 0.901 -> 0.820 | Hurt |
| mix 20% ground truth into NN affinities (ablation) | MWS · BSD500 test (MWS on interpolated affinities) | Rand index: 0.826 -> 0.878 | Improved |
| add 57% noise to NN affinities (ablation) | MWS · BSD500 test (MWS on interpolated affinities) | Rand index: 0.826 -> 0.813 | Hurt |
Closest papers by meaning
- The Mutex Watershed and its Objective: Efficient, Parameter-Free Graph Partitioning (TPAMI 2020 · similarity 0.98)
- Compassionately Conservative Balanced Cuts for Image Segmentation (CVPR 2018 · similarity 0.87)
- Learned Watershed: End-To-End Learning of Seeded Segmentation (ICCV 2017 · similarity 0.86)
- Dominant Sets for “Constrained” Image Segmentation (TPAMI 2018 · similarity 0.86)
- LoMix: Learnable Weighted Multi-Scale Logits Mixing for Medical Image Segmentation (NeurIPS 2025 · similarity 0.84)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.84)
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.
