Research map / Detection and segmentation / CNN features
Superpixel Sampling Networks
Varun Jampani, Deqing Sun, Ming-Yu Liu, Ming-Hsuan Yang, Jan Kautz
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 |
|---|---|---|---|
| replace SLIC superpixels with SSN_deep in bilateral inception network | DeepLab + BI · PascalVOC2012 test | IoU (%): 74.1 -> 75.3 | Improved |
| replace CRF with BI using SSN_deep | DeepLab · PascalVOC2012 test | IoU (%): 72.7 -> 75.3 | Improved |
| increase differentiable SLIC iterations v from 5 to 10 | SSN_deep (k=20) · Cityscapes 512x1024 | runtime (ms): 80 -> 101 | Hurt |
| increase feature dimension k from 10 to 20 | SSN_deep (v=10) · Cityscapes 512x1024 | runtime (ms): 90 -> 101 | Hurt |
| add learned deep features (SSN_deep instead of SSN_pix) (ablation) | SSN · Cityscapes 512x1024 | runtime (ms): 58 -> 101 | Hurt |
| add linear learned transform of XYLab features (SSN_linear instead of SSN_pix) (ablation) | SSN · BSDS500 | ASA / boundary recall | Improved |
| replace linear layer with deep network (SSN_deep over SSN_linear) (ablation) | SSN · BSDS500 | ASA / boundary recall | Improved |
| increase feature dimensionality k and iterations v (ablation) | SSN · BSDS500 | ASA / boundary recall | Improved |
| train on BSDS instead of PascalVOC for PascalVOC evaluation | SSN_deep · PascalVOC2012 | ASA | No clear effect |
| compare SSN_deep with SEAL superpixels | SSN_deep · Cityscapes | ASA / precision-recall | Mixed |
| relax nearest neighbor constraints in SLIC (SSN_pix vs SLIC) (ablation) | SSN_pix · BSDS500 | ASA / precision-recall | No clear effect |
Closest papers by meaning
- Superpixel Segmentation With Fully Convolutional Networks (CVPR 2020 · similarity 0.90)
- Learning the Superpixel in a Non-Iterative and Lifelong Manner (CVPR 2021 · similarity 0.89)
- Learning Superpixels With Segmentation-Aware Affinity Loss (CVPR 2018 · similarity 0.88)
- SPSN: Superpixel Prototype Sampling Network for RGB-D Salient Object Detection (ECCV 2022 · similarity 0.88)
- Differentiable Laplacian Matrix Guided Superpixel Segmentation (CVPR 2026 · similarity 0.88)
- Weakly Supervised Semantic Segmentation Using Superpixel Pooling Network (AAAI 2017 · 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.
