Research map / Detection and segmentation / Open-set and few-shot detection
Autoregressive Unsupervised Image Segmentation
Yassine Ouali, Céline Hudelot, Myriam Tami
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 |
|---|---|---|---|
| apply orderings via masking on low-level features (F2) instead of directly on inputs (F1) (ablation) | AC · Potsdam | accuracy: 39.3 -> 46.4 | Improved |
| add masking in the encoder (F2) versus unmasked encoder (F4) (ablation) | AC · Potsdam | accuracy: 35.1 -> 46.4 | Improved |
| use 8 orderings instead of 2 (ablation) | AC · Potsdam | accuracy: 43.2 -> 46.4 | Improved |
| use 8 orderings instead of 4 (ablation) | AC · Potsdam-3 | accuracy: 63.55 -> 66.4 | Improved |
| add a single shallow attention block (ablation) | AC raster-scan · Potsdam | accuracy: 45.2 -> 47.9 | Improved |
| combine raster-scan and zigzag orderings with attention (ablation) | AC · Potsdam | accuracy: 47.9 -> 49.3 | Improved |
| combine raster-scan and zigzag orderings with attention (ablation) | AC · Potsdam-3 | accuracy: 66.3 -> 65.4 | No clear effect |
| use hard ordering sampling instead of random (ablation) | AC · Potsdam | accuracy: 46.4 -> 48.9 | Improved |
| use hard ordering sampling instead of random (ablation) | AC · Potsdam-3 | accuracy: 66.4 -> 65.2 | Hurt |
| add rotation augmentation on the second forward pass (ablation) | AC · Potsdam | accuracy: 46.4 -> 48.5 | Improved |
| add rotation augmentation on the second forward pass (ablation) | AC · Potsdam-3 | accuracy: 66.4 -> 68.3 | Improved |
| add color jittering on the second forward pass (ablation) | AC · Potsdam-3 | accuracy: 66.4 -> 65.5 | Hurt |
Closest papers by meaning
- Segment Anything without Supervision (NeurIPS 2024 · similarity 0.88)
- Seg-VAR:Image Segmentation with Visual Autoregressive Modeling (NeurIPS 2025 · similarity 0.88)
- Unsupervised Universal Image Segmentation (CVPR 2024 · similarity 0.87)
- Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling (CVPR 2024 · similarity 0.86)
- Learn to Rectify the Bias of CLIP for Unsupervised Semantic Segmentation (CVPR 2024 · similarity 0.86)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.86)
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.
