Research map / Detection and segmentation / Open-set and few-shot detection
Evolved Part Masking for Self-Supervised Learning
Zhanzhou Feng, Shiliang Zhang
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 fixed masking with evolved grid-to-part masking | MAE, ViT-S · ImageNet-1K | Top-1 accuracy: 79.20 -> 79.89 | Improved |
| Replace fixed masking with evolved grid-to-part masking | MAE, ViT-S · ADE20K | mIoU: 38.81 -> 40.42 | Improved |
| Add evolved part masking | BEiT, ViT-S · ImageNet-1K | Top-1 accuracy: 79.05 -> 79.61 | Improved |
| Add evolved part masking | BEiT, ViT-S · ADE20K | mIoU: 38.39 -> 39.63 | Improved |
| Add evolved part masking | SimMIM, ViT-S · ImageNet-1K | Top-1 accuracy: 78.51 -> 79.08 | Improved |
| Add evolved part masking | SimMIM, ViT-S · ADE20K | mIoU: 38.13 -> 39.57 | Improved |
| Use evolved masking rather than fixed masking | ViT-S pre-training · ImageNet-1K | Top-1 accuracy: 79.77 -> 80.67 | Improved |
| Use evolved masking rather than fixed masking | ViT-S pre-training · ADE20K | mIoU: 41.81 -> 41.97 | Improved |
| Use evolved masking rather than fixed masking | ViT-S pre-training · COCO | AP-box: 38.91 -> 39.02 | Improved |
| Use dynamic α instead of fixed grid-mask weighting (α=0) (ablation) | MAE, ViT-S · ADE20K | mIoU: 36.67 -> 40.42 | Improved |
| Use dynamic α instead of fixed part-mask weighting (α=1) (ablation) | MAE, ViT-S · ImageNet-1K | Top-1 accuracy: 75.71 -> 79.89 | Improved |
| Use dynamic α instead of fixed part-mask weighting (α=1) (ablation) | MAE, ViT-S · ADE20K | mIoU: 35.82 -> 40.42 | Improved |
Closest papers by meaning
- Evolved Hierarchical Masking for Self-Supervised Learning (TPAMI 2024 · similarity 0.94)
- From Pose to Part: Weakly-Supervised Pose Evolution for Human Part Segmentation (TPAMI 2022 · similarity 0.87)
- Self-Guided Masked Autoencoder (NeurIPS 2024 · similarity 0.86)
- Unsupervised Part Segmentation Through Disentangling Appearance and Shape (CVPR 2021 · similarity 0.86)
- From Prototypes to General Distributions: An Efficient Curriculum for Masked Image Modeling (CVPR 2025 · similarity 0.85)
- Masking meets Supervision: A Strong Learning Alliance (CVPR 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.
