Research map / Detection and segmentation / Weakly supervised detection and segmentation
SEMPART: Self-supervised Multi-resolution Partitioning of Image Semantics
Sriram Ravindran, Debraj Basu
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 |
|---|---|---|---|
| add jointly optimized fine branch with coarse branch (ablation) | Sempart · DUT-OMRON saliency | mIoU: 0.662 -> 0.668 | Improved |
| add coarse branch to a fine-only model (ablation) | Sempart · DUT-OMRON saliency | mIoU: 0.645 -> 0.668 | Improved |
| add graph TV regularization on fine mask (ablation) | Sempart · DUT-OMRON saliency | mIoU: 0.637 -> 0.668 | Improved |
| add graph TV regularization on coarse mask (ablation) | Sempart · DUT-OMRON saliency | mIoU: 0.646 -> 0.668 | Improved |
| add graph TV coarse loss (ablation) | Sempart · DUTS-TE saliency | mIoU: 0.749 -> 0.749 | No clear effect |
| remove self-loops from affinity matrix (ablation) | Sempart · DUT-OMRON saliency | mIoU: 0.667 -> 0.668 | Improved |
| select foreground by least-corners rule (ablation) | Sempart · ECSSD saliency | mIoU: 0.825 -> 0.855 (vs framing prior) | Improved |
| apply bilateral filtering after inference (ablation) | Sempart-Fine · DUT-OMRON saliency | IoU: 0.668 -> 0.653 | Hurt |
| train SelfMask on Sempart masks | Sempart-Fine · DUT-OMRON saliency | IoU: 0.668 -> 0.698 | Improved |
| train SelfMask on Sempart masks | Sempart-Fine · ECSSD saliency | IoU: 0.855 -> 0.850 | Hurt |
| use smaller patch size (ViT-S/8 instead of ViT-S/16) in DINO backbone (ablation) | Sempart-Fine · DUT-OMRON saliency | IoU: 0.596 -> 0.668 | Improved |
| use fine mask instead of coarse mask (ablation) | Sempart · DUT-OMRON saliency | IoU: 0.640 -> 0.668 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MOVE — MOVE: Unsupervised Movable Object Segmentation and Detection (NeurIPS 2022) · better on 7 of 9 shared results
- TokenCut — TokenCut: Segmenting Objects in Images and Videos With Self-Supervised Transformer and Normalized Cut (TPAMI 2023) · better on 2 of 2 shared results
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
- Unsupervised Object Localization: Observing the Background To Discover Objects (CVPR 2023 · similarity 0.87)
- Unsupervised Semantic Segmentation with Self-supervised Object-centric Representations (ICLR 2023 · similarity 0.87)
- Self-Supervised Transformers for Unsupervised Object Discovery Using Normalized Cut (CVPR 2022 · similarity 0.87)
- Self-supervised co-salient object detection via feature correspondences at multiple scales (ECCV 2024 · similarity 0.86)
- ModuSeg: Decoupling Object Discovery and Semantic Retrieval for Training-Free Weakly Supervised Segmentation (ECCV 2026 · similarity 0.86)
- HEAP: Unsupervised Object Discovery and Localization with Contrastive Grouping (AAAI 2024 · similarity 0.85)
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.
