Research map / Detection and segmentation / CNN features
SAPA: Similarity-Aware Point Affiliation for Feature Upsampling
Hao Lu, Wenze Liu, Zixuan Ye, Hongtao Fu, Yuliang Liu, Zhiguo Cao
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 the upsampling stages with SAPA-G | SegFormer-B1 · ADE20K | mIoU: 42.82 -> 44.39 | Improved |
| Replace the upsampling stages with SAPA-G | MaskFormer-Swin-Base · ADE20K | mIoU: 53.53 -> 53.78 | Improved |
| Replace the upsampling stages with SAPA-G | Mask2Former-Swin-Base · ADE20K | mIoU: 53.94 -> 55.22 | Improved |
| Use SAPA-B instead of CARAFE | Faster R-CNN R50 · MS COCO | AP: 38.6 -> 37.8 | Hurt |
| Replace the upsampling stages with SAPA-G | BTS R50 · NYU Depth V2 | RMS: 0.419 -> 0.408 | Improved |
| Replace the upsampling stages with SAPA-G | A2U Matting R34 · Adobe Composition-1k test | SAD: 32.05 -> 30.98 | Improved |
| Add gated bilinear similarity rather than plain addition (ablation) | SAPA on SegFormer-B1 · ADE20K | mIoU: 43.2 -> 44.4 | Improved |
| Use exp normalization rather than no normalization (ablation) | SAPA on SegFormer-B1 · ADE20K | mIoU: 41.5 -> 43.2 | Improved |
| Use ReLU rather than exp normalization (ablation) | SAPA on SegFormer-B1 · ADE20K | mIoU: 43.2 -> 40.8 | Hurt |
| Use sigmoid rather than exp normalization (ablation) | SAPA on SegFormer-B1 · ADE20K | mIoU: 43.2 -> 42.8 | Hurt |
| Use softplus rather than exp normalization (ablation) | SAPA on SegFormer-B1 · ADE20K | mIoU: 43.2 -> 42.7 | Hurt |
| Increase kernel size from 5 to 7 (ablation) | SAPA on SegFormer-B1 · ADE20K | mIoU: 43.2 -> 42.5 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CARAFE — CARAFE: Content-Aware ReAssembly of FEatures (ICCV 2019) · better on 3 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Continuous Implicit Field with Local Distance Indicator for Arbitrary-Scale Point Cloud Upsampling (AAAI 2024 · similarity 0.84)
- P2Seg: Pointly-supervised Segmentation via Mutual Distillation (ICLR 2024 · similarity 0.84)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 2024 · similarity 0.84)
- Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling (CVPR 2024 · similarity 0.84)
- Semantic-aware SAM for Point-Prompted Instance Segmentation (CVPR 2024 · similarity 0.84)
- Semantic-Aware Transformation-Invariant RoI Align (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.
