Research map / Detection and segmentation / Weakly supervised detection and segmentation
Directed Acyclic Graph Continuous Max-Flow Image Segmentation for Unconstrained Label Orderings
John S. H. Baxter, Martin Rajchl, A. Jonathan McLeod, Jing Yun Yuan, Terry M. Peters
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 DAG label ordering with multiple parents instead of Potts | DAGMF · Synthetic Venn diagram | weighted DSC (%): 94.6% -> 95.4% | Improved |
| use DAG ordering instead of hierarchical HMF regularizing the square | DAGMF · Synthetic Venn diagram | weighted DSC (%): 95.0% -> 95.4% | Improved |
| use DAG ordering instead of hierarchical HMF regularizing the circle | DAGMF · Synthetic Venn diagram | weighted DSC (%): 94.9% -> 95.4% | Improved |
| group sGM with both V and WM via DAG instead of Potts | DAGMF · BrainWeb sGM | Dice coefficient (%): 76.6% -> 81.3% | Improved |
| DAG instead of HMF 1 for sGM | DAGMF · BrainWeb sGM | Dice coefficient (%): 76.1% -> 81.3% | Improved |
| DAG instead of HMF 1 for eCSF | DAGMF · BrainWeb eCSF | Dice coefficient (%): 76.5% -> 74.9% | Hurt |
| DAG instead of Potts for ventricles | DAGMF · BrainWeb V | Dice coefficient (%): 95.1% -> 95.7% | Improved |
| DAG instead of HMF 1 for ventricles | DAGMF · BrainWeb V | Dice coefficient (%): 96.1% -> 95.7% | Hurt |
| DAG instead of Potts on white matter | DAGMF · BrainWeb WM | Dice coefficient (%): 95.9% -> 95.9% | No clear effect |
| add constant regularization to data-term-only classifier (DAGMF) | DAGMF · Stanford indoor | accuracy rate (%): 84.4% -> 85.9% | Improved |
| DAGMF instead of HMF/Potts models | DAGMF · Stanford indoor | accuracy rate (%): 85.7% -> 85.9% | No clear effect |
| cyclic DAGMF hue regularization instead of Potts | DAGMF (N=36, M=16) · Synthetic hue reconstruction | hue error: 64% -> 79% | Improved |
Closest papers by meaning
- Consistent estimation of the max-flow problem: Towards unsupervised image segmentation (TPAMI 2020 · similarity 0.83)
- MRF Optimization with Separable Convex Prior on Partially Ordered Labels (ECCV 2018 · similarity 0.83)
- Topograph: An Efficient Graph-Based Framework for Strictly Topology Preserving Image Segmentation (ICLR 2025 · similarity 0.81)
- A Geometric Approach to Image Labeling (ECCV 2016 · similarity 0.81)
- Soft Self-labeling and Potts Relaxations for Weakly-supervised Segmentation (CVPR 2025 · similarity 0.80)
- HGDL: Heterogeneous Graph Label Distribution Learning (NeurIPS 2024 · similarity 0.79)
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.
