Research map / Architectures and efficiency / Attention and transformers
SMMix: Self-Motivated Image Mixing for Vision Transformers
Mengzhao Chen, Mingbao Lin, Zhihang Lin, Yuxin Zhang, Fei Chao, Rongrong Ji
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 CutMix with SMMix | DeiT-T/S/B, CaiT, PVT, Swin-T · ImageNet-1k val | top-1 accuracy: 79.8 -> 81.1 (DeiT-S); 81.2 -> 81.8 (Swin-T) | Improved |
| add max-min attention image mixing (ablation) | DeiT-S · ImageNet-1k val | top-1 accuracy: 79.8 -> 80.4 | Improved |
| add fine-grained label assignment to max-min mixing (ablation) | DeiT-S · ImageNet-1k val | top-1 accuracy: 80.4 -> 80.9 | Improved |
| add feature consistency constraint (ablation) | SMMix DeiT-S · ImageNet-1k val | top-1 accuracy: 80.8 -> 81.1 | Improved |
| add TransMix or TokenMix label reconstruction to SMMix (ablation) | SMMix DeiT-S · ImageNet-1k val | top-1 accuracy: 81.1 -> 81.1 (TransMix); 81.1 -> 81.2 (TokenMix) | No clear effect |
| use pre-trained same-size DeiT-T as motivated model instead of self (ablation) | SMMix DeiT-T · ImageNet-1k val | top-1 accuracy: 73.6 -> 73.5 | No clear effect |
| use larger pre-trained DeiT-S as motivated model (ablation) | SMMix DeiT-T · ImageNet-1k val | top-1 accuracy: 73.6 -> 74.1 | Improved |
| vary side ratio delta (fixed 0.5, U(0,1), U(0.25,0.75)) (ablation) | SMMix DeiT-S · ImageNet-1k val | top-1 accuracy: 81.0 / 81.0 / 81.1 | No clear effect |
| pretrain with SMMix instead of CutMix before transfer | PVT-T/S · ADE20K and COCO | mIoU / box AP: 41.9 -> 43.0 mIoU (PVT-S); 40.4 -> 41.0 AP (PVT-S) | Improved |
| train with SMMix instead of CutMix | DeiT-S · ImageNet-A | top-1 accuracy: 18.7 -> 22.0 | Improved |
| use SMMix instead of CutMix | mixed image quality · ImageNet-1k mixed images | top-2 accuracy of mixed images: 23.8% -> 48.3% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- TokenMix — TokenMix: Rethinking Image Mixing for Data Augmentation in Vision Transformers (ECCV 2022) · better on 1 of 2 shared results
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
- TokenMix: Rethinking Image Mixing for Data Augmentation in Vision Transformers (ECCV 2022 · similarity 0.91)
- TransMix: Attend To Mix for Vision Transformers (CVPR 2022 · similarity 0.91)
- MixPro: Data Augmentation with MaskMix and Progressive Attention Labeling for Vision Transformer (ICLR 2023 · similarity 0.90)
- TiMix: Text-Aware Image Mixing for Effective Vision-Language Pre-training (AAAI 2024 · similarity 0.87)
- DiffuseMix: Label-Preserving Data Augmentation with Diffusion Models (CVPR 2024 · similarity 0.86)
- Supervision Interpolation via LossMix: Generalizing Mixup for Object Detection and Beyond (AAAI 2024 · 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.
