Research map / Detection and segmentation / Weakly supervised detection and segmentation
ScribbleSup: Scribble-Supervised Convolutional Networks for 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 pairwise terms for propagating labels between superpixels (ablation) | ScribbleSup · PASCAL VOC 2012 validation | mIoU: 60.5 -> 63.1 | Helped |
| jointly propagate labels and train the network instead of first generating masks with LazySnapping | scribble-supervised segmentation · PASCAL VOC 2012 validation | mIoU: 53.8 -> 63.1 | Helped |
| shorten scribbles to length ratio 0.8 (ablation) | ScribbleSup · PASCAL VOC 2012 validation | mIoU: 63.1 -> 61.8 | Hurt |
| shorten scribbles to length ratio 0.5 (ablation) | ScribbleSup · PASCAL VOC 2012 validation | mIoU: 63.1 -> 58.5 | Hurt |
| shorten scribbles to length ratio 0.3 (ablation) | ScribbleSup · PASCAL VOC 2012 validation | mIoU: 63.1 -> 54.3 | Hurt |
| replace scribbles with spots (ablation) | ScribbleSup · PASCAL VOC 2012 validation | mIoU: 63.1 -> 51.6 | Hurt |
| replace mask annotations with scribbles for the same 11k training images (ablation) | ScribbleSup · PASCAL VOC 2012 validation | mIoU: 68.5 -> 63.1 | Hurt |
| add 10k VOC 2007 scribble-labeled images to the 11k mask-labeled training images | ScribbleSup · PASCAL VOC 2012 validation | mIoU: 68.5 -> 71.3 | Helped |
| add 10k VOC 2007 scribble-labeled images | ScribbleSup · PASCAL-CONTEXT validation | mIoU: 36.1 -> 39.3 | Helped |
| add VOC 2007 scribble-labeled images to the mask-labeled training set | ScribbleSup · PASCAL-CONTEXT validation | mIoU: 37.7 -> 42.0 | Helped |
| replace mask annotations with scribbles on the 5k PASCAL-CONTEXT training images (ablation) | ScribbleSup · PASCAL-CONTEXT validation | mIoU: 37.7 -> 36.1 | Hurt |
Closest papers by meaning
- Scribbles for All: Benchmarking Scribble Supervised Segmentation Across Datasets (NeurIPS 2024 · similarity 0.94)
- Scribble-Supervised Semantic Segmentation with Prototype-based Feature Augmentation (ICML 2024 · similarity 0.93)
- Scribble Hides Class: Promoting Scribble-Based Weakly-Supervised Semantic Segmentation with Its Class Label (AAAI 2024 · similarity 0.92)
- SAM-CP: Marrying SAM with Composable Prompts for Versatile Segmentation (ICLR 2025 · similarity 0.85)
- Exploring Simple Open-Vocabulary Semantic Segmentation (CVPR 2025 · similarity 0.85)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 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.