Research map / Detection and segmentation / CNN features
Contextrast++: Robust Multi-Scale Contextual Contrastive Learning for Semantic Segmentation
Changki Sung, Hyungtae Lim, Wanhee Kim, Youngwoo Seo, Hyun Myung
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 |
|---|---|---|---|
| add adaptive fusion module (ablation) | Contextrast++ HRNet · Cityscapes val (HRNet) | mIoU: 81.88 -> 82.18 | Improved |
| add L_AA with memory bank (ablation) | Contextrast++ HRNet · Cityscapes val (HRNet) | mIoU: 82.18 -> 82.35 | Improved |
| use BANE sampling instead of semi-hard sampling (ablation) | Contextrast++ HRNet · Cityscapes val (HRNet) | mIoU: 82.23 -> 82.44 | Improved |
| combine adaptive fusion, L_AA and BANE (ablation) | Contextrast++ HRNet · Cityscapes val (HRNet) | mIoU: 81.88 -> 82.87 | Improved |
| aggregate all four scales (vs first layer only) (ablation) | Contextrast++ HRNet · Cityscapes val (HRNet) | mIoU: 81.02 -> 82.87 | Improved |
| use self-attention + MLP fusion instead of gated channel-wise fusion (ablation) | Contextrast++ HRNet · Cityscapes val (HRNet) | mIoU: 80.46 -> 82.87 | Improved |
| use deepest-layer class-prioritized anchors instead of earliest-layer (ablation) | Contextrast++ HRNet · Cityscapes val (HRNet) | mIoU: 81.73 -> 82.87 | Improved |
| use anchor memory instead of pixel memory (ablation) | Contextrast++ HRNet · Cityscapes val (HRNet) | mIoU: 81.94 -> 82.87 | Improved |
| store class-prioritized anchors instead of all-layer anchors (ablation) | Contextrast++ HRNet · Cityscapes val (HRNet) | mIoU: 80.00 -> 82.87 | Improved |
| use Contextrast++ vs Contextrast on tail classes | HRNet · PASCAL-Context | tail-class IoU: 28.35 -> 32.36 | Improved |
| apply Contextrast++ contrastive training to Mask2Former | Mask2Former Swin-T · Cityscapes val | mIoU: 81.87 -> 82.92 | Improved |
| apply Contextrast++ to SegFormer MiT-B0 | SegFormer · Cityscapes val | mIoU: 78.22 -> 78.55 | No clear effect |
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.
- MobileNetV2 — MobileNetV2: Inverted Residuals and Linear Bottlenecks (CVPR 2018) · better on 1 of 1 shared result
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
- Contextrast: Contextual Contrastive Learning for Semantic Segmentation (CVPR 2024 · similarity 0.96)
- Multi-Scale and Cross-Scale Contrastive Learning for Semantic Segmentation (ECCV 2022 · similarity 0.89)
- ISNet: Integrate Image-Level and Semantic-Level Context for Semantic Segmentation (ICCV 2021 · similarity 0.88)
- Exploring Cross-Image Pixel Contrast for Semantic Segmentation (ICCV 2021 · similarity 0.88)
- Multi-Scale Context Intertwining for Semantic Segmentation (ECCV 2018 · similarity 0.88)
- Region-Aware Contrastive Learning for Semantic Segmentation (ICCV 2021 · similarity 0.88)
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.
