Research map / Detection and segmentation / Weakly supervised detection and segmentation
Tree Energy Loss: Towards Sparsely Annotated Semantic Segmentation
Check 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 |
|---|---|---|---|
| Add TEL to DeepLabV3+ | DeepLabV3+ with partial cross-entropy · Pascal VOC 2012, point supervision | mIoU: 58.5% -> 64.9% | Helped |
| Replace Seminar's DenseCRF loss with TEL | Seminar · Pascal VOC 2012, point supervision | mIoU: +1.7% | Helped |
| Use TEL instead of A²GNN | Scribble-supervised segmentation · Pascal VOC 2012, scribble supervision | mIoU: +1.1% | Helped |
| Add TEL rather than DenseCRF Loss | Block-supervised segmentation · Cityscapes and ADE20k, 10%, 20%, and 50% labels | mIoU | Helped |
| Use L1 distance for pseudo-label assignment rather than cross-entropy, dot product, or L2 distance (ablation) | TEL · Pascal VOC 2012 val, scribble supervision | mIoU: 76.0, 76.6, 75.1 -> 77.1 | Helped |
| Add low-level affinity (ablation) | TEL · Pascal VOC 2012 val, scribble supervision | mIoU: +7.5 | Helped |
| Add high-level affinity (ablation) | TEL · Pascal VOC 2012 val, scribble supervision | mIoU: +3.1 | Helped |
| Use both low- and high-level affinities (ablation) | TEL · Pascal VOC 2012 val, scribble supervision | mIoU: +8.3 | Helped |
| Generate low-level affinity with MST instead of Bilateral Filter (ablation) | TEL · Pascal VOC 2012 val, scribble supervision | mIoU: +1.3 | Helped |
| Generate high-level affinity with MST instead of Non-Local operation (ablation) | TEL · Pascal VOC 2012 val, scribble supervision | mIoU: +1.7 | Helped |
| Apply low-level affinity before high-level affinity rather than in parallel or high-level first (ablation) | TEL · Pascal VOC 2012 val, scribble supervision | mIoU: LH-P 76.4; HL-C 75.8; LH-C 77.1 | Helped |
| Set loss balance λ to 0.4 rather than the tested alternatives (ablation) | TEL · Pascal VOC 2012 val, scribble supervision | mIoU: 74.9, 76.0, 76.4, 77.1, 77.0 | Helped |
Closest papers by meaning
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.86)
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (AAAI 2024 · similarity 0.86)
- Soft Self-labeling and Potts Relaxations for Weakly-supervised Segmentation (CVPR 2025 · similarity 0.86)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.86)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.85)
- Point Cloud Semantic Segmentation with Sparse and Inhomogeneous Annotations (AAAI 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.