Research map / Detection and segmentation / CNN features
MixPro: Data Augmentation with MaskMix and Progressive Attention Labeling for Vision Transformer
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 |
|---|---|---|---|
| Apply MixPro instead of TransMix | DeiT-T · ImageNet-1K | Top-1 accuracy (%): 72.6 -> 73.8 | Helped |
| Apply MixPro instead of TransMix | DeiT-S · ImageNet-1K | Top-1 accuracy (%): 80.7 -> 81.3 | Helped |
| Apply MixPro pre-training | DeiT-S with Linear decoder · Pascal Context semantic segmentation | mIoU: 49.7 -> 50.3 | Helped |
| Apply MixPro pre-training | PVT-S · COCO val2017 object detection | AP^b: 40.9 -> 41.4 | Helped |
| Apply MixPro pre-training | DeiT-S · ImageNet-A | Top1-Acc: 21.1% -> 22.4 | Helped |
| Add PAL to MaskMix (ablation) | DeiT-S with MaskMix · ImageNet-1K | Top-1 accuracy (%): 81.0 -> 81.3 | Helped |
| Add MaskMix instead of CutMix (ablation) | DeiT-S with PAL · ImageNet-1K | Top-1 accuracy (%): 81.1 -> 81.3 | Helped |
| Use random mask patches at 4x scale instead of region masks (ablation) | DeiT-T · ImageNet-1K | Top-1 accuracy (%): 72.2 -> 73.2 | Helped |
| Use random mask patches at 4x scale instead of region masks (ablation) | DeiT-S · ImageNet-1K | Top-1 accuracy (%): 79.8 -> 80.8 | Helped |
| Align mask borders with image patch borders (ablation) | TransMix on DeiT-T · ImageNet-1K | Top-1 accuracy (%): 72.6 -> 73.1 | Helped |
| Use PAL's cosine-distance progressive factor instead of alternative alpha strategies (ablation) | DeiT-T · ImageNet-1K | Top-1 error (%): 27.3 -> 26.2 | Helped |
| Randomly shuffle cropped patches (ablation) | DeiT-T · ImageNet | results: 73.2 -> 69.8 | Hurt |
Closest papers by meaning
- TdAttenMix: Top-Down Attention Guided Mixup (AAAI 2025 · similarity 0.87)
- DiffuseMix: Label-Preserving Data Augmentation with Diffusion Models (CVPR 2024 · similarity 0.86)
- Configuring Data Augmentations to Reduce Variance Shift in Positional Embedding of Vision Transformers (AAAI 2025 · similarity 0.85)
- Skip-Attention: Improving Vision Transformers by Paying Less Attention (ICLR 2024 · similarity 0.85)
- Enhance Image Classification via Inter-Class Image Mixup with Diffusion Model (CVPR 2024 · similarity 0.84)
- LookHere: Vision Transformers with Directed Attention Generalize and Extrapolate (NeurIPS 2024 · similarity 0.84)
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.