Research map / Detection and segmentation / Weakly supervised detection and segmentation
Exploratory Inference Learning for Scribble Supervised Semantic Segmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 exploratory inference learning to select confident pseudo-labels | Scribble-supervised segmentation · PASCAL VOC 2012 validation | mIoU: 76.2 -> 77.9 | Helped |
| Use exploratory inference learning to select confident pseudo-labels | Scribble-supervised segmentation · PASCAL Context validation | mIoU: 43.1 -> 43.8 | Helped |
| Add random-policy selection with a policy assessor (ablation) | EIL segmenter · PASCAL VOC 2012 validation, subset | mIoU: 71.1 -> 72.2 | Helped |
| Add random-policy selection with a policy assessor (ablation) | EIL segmenter · PASCAL VOC 2012 validation, full set | mIoU: 73.5 -> 74.8 | Helped |
| Replace random inference policies with trained policy-searcher actions (ablation) | EIL segmenter · PASCAL VOC 2012 validation, subset | mIoU: 71.1 -> 50.3 | Hurt |
| Replace random inference policies with trained policy-searcher actions (ablation) | EIL segmenter · PASCAL VOC 2012 validation, full set | mIoU: 73.5 -> 51.7 | Hurt |
| Add joint training of the policy searcher and policy assessor (ablation) | EIL segmenter · PASCAL VOC 2012 validation, subset | mIoU: 72.2 -> 76.2 | Helped |
| Add joint training of the policy searcher and policy assessor (ablation) | EIL segmenter · PASCAL VOC 2012 validation, full set | mIoU: 74.8 -> 77.9 | Helped |
| Extend pseudo-label inference radius beyond 101 (ablation) | EIL segmenter · PASCAL VOC 2012 validation | mIoU | Hurt |
| Use different feature-distance functions in the policy reward (ablation) | EIL reward criterion · PASCAL VOC 2012 validation | mIoU | No clear effect |
| Use one-dimensional rather than two-dimensional policy actions (ablation) | EIL policy searcher · PASCAL VOC 2012 validation | mIoU: 76.2 -> 75.0; 77.9 -> 76.3 | Hurt |
| Use C-dimensional rather than two-dimensional policy actions (ablation) | EIL policy searcher · PASCAL VOC 2012 validation | mIoU: 76.2 -> 75.3; 77.9 -> 76.8 | Hurt |
Closest papers by meaning
- Scribble Hides Class: Promoting Scribble-Based Weakly-Supervised Semantic Segmentation with Its Class Label (AAAI 2024 · similarity 0.90)
- Scribbles for All: Benchmarking Scribble Supervised Segmentation Across Datasets (NeurIPS 2024 · similarity 0.89)
- Scribble-Supervised Semantic Segmentation with Prototype-based Feature Augmentation (ICML 2024 · similarity 0.87)
- Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation (NeurIPS 2025 · similarity 0.86)
- Seg-ReSearch: Segmentation with Interleaved Reasoning and External Search (ICML 2026 · similarity 0.86)
- SAM-R1: Leveraging SAM for Reward Feedback in Multimodal Segmentation via Reinforcement Learning (NeurIPS 2025 · 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.