Research map / Detection and segmentation / Open-set and few-shot detection
S2-FracMix: Self-Saliency Fractal Mixup
Khawar Islam, Arif Mahmood, Xin Jin, NAVEED AKHTAR
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 AdAutoMix with S2-FracMix | ResNet-18 · CIFAR-100 | Top-1 accuracy (%): 82.32 -> 82.74 | Improved |
| replace AdAutoMix with S2-FracMix | ResNeXt-50 · Tiny-ImageNet | Top-1 accuracy (%): 72.89 -> 74.27 | Improved |
| replace AdAutoMix with S2-FracMix | ResNet-50 · ImageNet-1K | Top-1 accuracy (%): 78.04 -> 78.54 | Improved |
| replace AdAutoMix with S2-FracMix | ResNet-18 · CIFAR-100-C | Corruption accuracy (%): 51.44 -> 53.84 | Improved |
| replace AdAutoMix with S2-FracMix | ResNet-18 · CIFAR-100 FGSM | FGSM error (%): 75.66 -> 72.52 | Improved |
| replace AdAutoMix with S2-FracMix | ResNet-18 · Stanford-Cars | accuracy (%): 89.19 -> 90.56 | Improved |
| use self-saliency (A+A) instead of cross-image saliency (A+B) (ablation) | Saliency mixing · CIFAR-100 (ResNet-18) | Top-1 accuracy (%): 79.12 -> 79.54 | Improved |
| add FracMix fractal mixing to saliency mixing (ablation) | S2 mixing · CIFAR-100 (ResNet-18) | Top-1 accuracy (%): 79.54 -> 82.74 | Improved |
| add S2 self-saliency to FracMix (ablation) | FracMix · CIFAR-100 (ResNeXt-50) | Top-1 accuracy (%): 82.22 -> 84.91 | Improved |
| use S2-FracMix instead of FMix in high-level mixing (ablation) | Mixup+CutMix+ResizeMix · CIFAR-100 (ResNet-18) | Top-1 accuracy (%): 80.24 -> 82.74 | Improved |
| add saliency weighting (ablation) | S2-FracMix · Stanford-Cars | Top-1 accuracy (%): 91.87 -> 92.78 | Improved |
| apply fractals locally to salient regions instead of globally (ablation) | S2-FracMix · Stanford-Cars | Top-1 accuracy (%): 92.27 -> 92.78 | Improved |
Closest papers by meaning
- Saliency Grafting: Innocuous Attribution-Guided Mixup with Calibrated Label Mixing (AAAI 2022 · similarity 0.87)
- SaliencyMix: A Saliency Guided Data Augmentation Strategy for Better Regularization (ICLR 2021 · similarity 0.87)
- Model-Agnostic and Efficient Mixup Augmentation Guided by Saliency Maps (TPAMI 2025 · similarity 0.86)
- Saliency Benchmarking Made Easy: Separating Models, Maps and Metrics (ECCV 2018 · similarity 0.86)
- GuidedMixup: An Efficient Mixup Strategy Guided by Saliency Maps (AAAI 2023 · similarity 0.86)
- Advancing Saliency Ranking with Human Fixations: Dataset Models and Benchmarks (CVPR 2024 · similarity 0.83)
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.
