Research map / Architectures and efficiency / RNNs and normalization
A Unified Analysis of Mixed Sample Data Augmentation: A Loss Function Perspective
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 |
|---|---|---|---|
| use Mixup vs CutMix when objects are small (large crop) | MSDA on ImageNet-100 64x64 · Scenario 1: large crop | accuracy (%): 58.3 -> 64.4 | Helped |
| use Mixup vs CutMix when objects are large (small crop) | MSDA on ImageNet-100 64x64 · Scenario 2: small crop | accuracy (%): 67.0 -> 67.7 | Helped |
| use HMix instead of CutMix | ResNet-50 · ImageNet-1K | top-1 accuracy (%): 78.04 -> 78.38 | Helped |
| use GMix instead of CutMix | ResNet-50 · ImageNet-1K | top-1 accuracy (%): 78.04 -> 78.13 | Helped |
| use HMix instead of Stochastic Mixup & CutMix | ResNet-50 · ImageNet-1K | top-1 accuracy (%): 78.13 -> 78.38 | Helped |
| use GMix instead of CutMix | RN56 · CIFAR-100 | accuracy (%): 74.83 -> 75.75 | Helped |
| use HMix or GMix instead of Stochastic Mixup & CutMix | CIFAR-100 networks · CIFAR-100 | accuracy (%): 79.25 -> 79.25 (PreActRN18) | No clear effect |
| use HMix/GMix instead of data-aware PuzzleMix | CIFAR-100 networks · CIFAR-100 | accuracy (%): 76.51 vs 76.15 (WRN28-2); 82.46 vs 81.45 (PreActRN50) | Mixed |
| use Mixup instead of no augmentation | RN56 · CIFAR-100 | accuracy (%): 73.23 -> 73.12 | No clear effect |
Closest papers by meaning
- Adversarial AutoMixup (ICLR 2024 · similarity 0.91)
- Harnessing Hard Mixed Samples with Decoupled Regularizer (NeurIPS 2023 · similarity 0.89)
- StyleMix: Separating Content and Style for Enhanced Data Augmentation (CVPR 2021 · similarity 0.88)
- Noisy Feature Mixup (ICLR 2022 · similarity 0.88)
- Catch-Up Mix: Catch-Up Class for Struggling Filters in CNN (AAAI 2024 · similarity 0.87)
- First-Order Manifold Data Augmentation for Regression Learning (ICML 2024 · similarity 0.86)
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.