Research map / Detection and segmentation / CNN features
Crafting Better Contrastive Views for Siamese Representation Learning
Check 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 |
|---|---|---|---|
| Replace RandomCrop with ContrastiveCrop | SimCLR, MoCo, BYOL, and SimSiam · CIFAR-10, CIFAR-100, Tiny ImageNet, and STL-10 | top-1 accuracy: 0.4% ~ 2.0% | Helped |
| Replace RandomCrop with ContrastiveCrop | SimCLR, MoCo V1, MoCo V2, and SimSiam · IN-1K | top-1 accuracy: 0.25%, 1.09%, 0.49% and 0.33% | Helped |
| Replace RandomCrop with ContrastiveCrop | MoCo V1 and MoCo V2 · VOC detection, COCO detection, and COCO instance segmentation | AP | Helped |
| Add semantic-aware localization to RandomCrop (ablation) | MoCo V2 · IN-200 | top-1 accuracy | Helped |
| Increase localization threshold k above 0.25 (ablation) | MoCo V2 with semantic-aware localization · IN-200 | top-1 accuracy | Hurt |
| Set localization-box update frequency to 10%, 20%, or 30% instead of 0% (ablation) | MoCo V2 · IN-200 | top-1 accuracy: 63.43 -> 64.40; 63.43 -> 64.61; 63.43 -> 64.40 | Helped |
| Set localization-box update frequency to 50% instead of 0% (ablation) | MoCo V2 · IN-200 | top-1 accuracy: 63.43 -> 64.11 | Helped |
| Add center-suppressed sampling with alpha < 1 (ablation) | MoCo V2 with semantic-aware localization · IN-200 | top-1 accuracy | Helped |
| Use center-concentrated sampling with alpha > 1 (ablation) | MoCo V2 with semantic-aware localization · IN-200 | top-1 accuracy | Hurt |
| Replace RandomCrop with ContrastiveCrop while retaining other transformations (ablation) | MoCo V2 · IN-200 | top-1 accuracy: 0.3% ~ 0.8%; 1.2% | Helped |
| Replace RandomCrop with ContrastiveCrop without other transformations (ablation) | MoCo V2 · IN-200 | top-1 accuracy: 45.5 -> 45.9 | Helped |
Closest papers by meaning
- Implicit Contrastive Representation Learning with Guided Stop-gradient (NeurIPS 2023 · similarity 0.89)
- A Simple Framework for Contrastive Learning of Visual Representations (ICML 2020 · similarity 0.88)
- Revisiting Contrastive Methods for Unsupervised Learning of Visual Representations (NeurIPS 2021 · similarity 0.87)
- Exploring Simple Siamese Representation Learning (CVPR 2021 · similarity 0.87)
- Enhancing Contrastive Learning Inspired by the Philosophy of “The Blind Men and the Elephant” (AAAI 2025 · similarity 0.86)
- Diversity Is All You Need for Contrastive Learning: Spectral Bounds on Gradient Magnitudes (NeurIPS 2025 · similarity 0.85)
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.