Research map / Label-efficient and robust learning / Self-supervised and contrastive
Can Semantic Labels Assist Self-Supervised Visual Representation Learning?
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 |
|---|---|---|---|
| Use SCAN's semantic-and-appearance positive neighbors instead of MoCo-v2 pretraining | ResNet-50 pretraining · VOC semantic segmentation | mIoU: 74.4 -> 76.6 | Helped |
| Use SCAN's semantic-and-appearance positive neighbors instead of MoCo-v2 pretraining | ResNet-50 FPN pretraining · COCO detection, 1× schedule | APbb: 39.2 -> 40.9 | Helped |
| Use SCAN's semantic-and-appearance positive neighbors instead of MoCo-v2 pretraining | ResNet-50 FPN pretraining · LVIS-v0.5 instance segmentation | AP: 25.3 -> 25.4 | Helped |
| Use SCAN's semantic-and-appearance positive neighbors instead of MoCo-v2 pretraining | ResNet-50 pretraining · ImageNet-1K linear evaluation | Top-1 accuracy: +3.9% | Helped |
| Add two top-ranked images sharing both semantic and appearance similarity as positive neighbors (ablation) | SCAN · COCO detection, 1× schedule | APbb: 39.2% to 40.9% | Helped |
| Increase positive-neighbor count from 2 to 8, including appearance-dissimilar images of the same class (ablation) | SCAN · COCO instance segmentation, 1× schedule | APmk: 37.2% to 36.3% | Hurt |
| Use semantic labels through SCAN rather than MoCo-v2 pretraining | ResNet-50 C4 pretraining · COCO detection, small objects | APbb S: 23.8 -> 24.3 | Helped |
| Use semantic labels through SCAN rather than MoCo-v2 pretraining | ResNet-50 C4 pretraining · COCO detection, large objects | APbb L: 56.5 -> 56.3 | Hurt |
Closest papers by meaning
- A Theory-Driven Self-Labeling Refinement Method for Contrastive Representation Learning (NeurIPS 2021 · similarity 0.90)
- When Does Contrastive Visual Representation Learning Work? (CVPR 2022 · similarity 0.89)
- Revisiting Self-Supervised Visual Representation Learning (CVPR 2019 · similarity 0.89)
- An Augmentation-Aware Theory for Self-Supervised Contrastive Learning (ICML 2025 · similarity 0.88)
- Can Generative Models Improve Self-Supervised Representation Learning? (AAAI 2025 · similarity 0.87)
- Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning (NeurIPS 2025 · similarity 0.87)
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.