Research map / Detection and segmentation / CNN features
Learning Superpixels With Segmentation-Aware Affinity Loss
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 |
|---|---|---|---|
| replace ERS superpixels with SEAL-ERS superpixels | BI semantic segmentation · PascalVOC 2012 test | IoU: 74.5 -> 75.4 | Helped |
| replace ERS superpixels with SEAL-ERS superpixels | BI semantic segmentation · PascalVOC 2012 test | IoU: 75.0 with 600 superpixels; 75.4 with 1000 | Helped |
| replace SLIC superpixels with SEAL-ERS superpixels | SO salient object detection · ECSSD | MAE: 0.1719 -> 0.1633 | Helped |
| replace SLIC superpixels with SEAL-ERS superpixels | SO salient object detection · ECSSD | Weighted Fβ: 0.5101 -> 0.5237 | Helped |
| replace SLIC superpixels with SEAL-ERS superpixels | GMR salient object detection · ECSSD | MAE: 0.1892 -> 0.1841 | Helped |
| replace SLIC superpixels with SEAL-ERS superpixels | GMR salient object detection · ECSSD | Weighted Fβ: 0.4909 -> 0.4955 | Helped |
| use VGG16 pre-trained features instead of RGB features for pixel affinities | ERS superpixel segmentation · BSDS500 test | ASA | Hurt |
| derive pixel affinities from HED edge predictions | ERS superpixel segmentation · BSDS500 test | Boundary Recall | Hurt |
| train affinities with class-balancing BCE instead of SEAL | PAN with ERS · BSDS500 test | ASA | Hurt |
| add segmentation-aware weighting of boundary-pixel BCE by over-segmentation error | PAN affinity learning with ERS · BSDS500 test | ASA | Helped |
| remove one-way affinity prediction (ablation) | PAN · BSDS500 test | ASA | No clear effect |
| replace first-layer 1×7 filters with 7×7 filters (ablation) | PAN · BSDS500 test | ASA | No clear effect |
Closest papers by meaning
- Superpixel Segmentation With Fully Convolutional Networks (CVPR 2020 · similarity 0.91)
- Learning the Superpixel in a Non-Iterative and Lifelong Manner (CVPR 2021 · similarity 0.88)
- Differentiable Laplacian Matrix Guided Superpixel Segmentation (CVPR 2026 · similarity 0.87)
- Semantic-Aware Superpixel for Weakly Supervised Semantic Segmentation (AAAI 2023 · similarity 0.86)
- Hierarchical Histogram Threshold Segmentation - Auto-terminating High-detail Oversegmentation (CVPR 2024 · similarity 0.85)
- Learning Invariant Inter-pixel Correlations for Superpixel Generation (AAAI 2024 · similarity 0.84)
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.