Research map / Label-efficient and robust learning / Self-supervised and contrastive
RegionCL: Exploring Contrastive Region Pairs for Self-Supervised Representation Learning
Yufei Xu, Qiming Zhang, Jing Zhang, Dacheng Tao
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 RegionCL region-level contrastive pairs | MoCov2 · ImageNet linear classification | Top-1: 67.5 -> 69.4 | Improved |
| add RegionCL region-level contrastive pairs | DenseCL · ImageNet linear classification | Top-1: 63.6 -> 68.5 | Improved |
| add RegionCL region-level contrastive pairs | SimSiam · ImageNet linear classification | Top-1: 68.1 -> 71.3 | Improved |
| add RegionCL to MoBy | MoBy Swin-T · ImageNet linear (100 ep) | Top-1: 70.9 -> 73.9 | Improved |
| add RegionCL to MoBy | MoBy Swin-T · ImageNet linear (300 ep) | Top-1: 75.3 -> 77.0 | Improved |
| add RegionCL with 1% labels | DenseCL · ImageNet 1% labels | Top-1: 38.9 -> 47.8 | Improved |
| add RegionCL | MoCov2 · COCO Mask-RCNN FPN 1x | APbb: 38.9 -> 40.1 | Improved |
| add RegionCL | DenseCL · COCO RetinaNet 1x | APbb: 37.7 -> 38.8 | Improved |
| add RegionCL | MoCov2 · Cityscapes instance segmentation | AP: 33.9 -> 34.9 | Improved |
| use paste view positives (Eq. 2 term) (ablation) | RegionCL-M · ImageNet 20-NN (100 ep) | accuracy: 49.3 -> 50.0 | Improved |
| use canvas view positives (ablation) | RegionCL-M · ImageNet 20-NN (100 ep) | accuracy: 49.3 -> 51.8 | Improved |
| use both paste and canvas views (ablation) | RegionCL-M · ImageNet 20-NN (100 ep) | accuracy: 49.3 -> 54.6 | Improved |
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.
- UnMix — Un-mix: Rethinking Image Mixtures for Unsupervised Visual Representation Learning (AAAI 2022) · 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
- Revisiting Contrastive Methods for Unsupervised Learning of Visual Representations (NeurIPS 2021 · similarity 0.89)
- Region-Aware Contrastive Learning for Semantic Segmentation (ICCV 2021 · similarity 0.89)
- FineCLIP: Self-distilled Region-based CLIP for Better Fine-grained Understanding (NeurIPS 2024 · similarity 0.88)
- REN: Fast and Efficient Region Encodings from Patch-Based Image Encoders (NeurIPS 2025 · similarity 0.86)
- Linear Differential Vision Transformer: Learning Visual Contrasts via Pairwise Differentials (NeurIPS 2025 · similarity 0.86)
- CLIM: Contrastive Language-Image Mosaic for Region Representation (AAAI 2024 · 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.
