Research map / Label-efficient and robust learning / Recognition and long-tail
SAFA: Sample-Adaptive Feature Augmentation for Long-Tailed Image Classification
Yan Hong, Jianfu Zhang, Zhongyi Sun, Ke Yan
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 |
|---|---|---|---|
| add cycle reconstruction loss L_r (ablation) | LDAM-DRW-SAFA · CIFAR-LT-100 rho=200 | top-1 error: 64.87 -> 57.53 | Improved |
| add cycle reconstruction loss L_r (ablation) | LDAM-DRW-SAFA · CIFAR-LT-10 rho=200 | top-1 error: 31.21 -> 22.47 | Improved |
| add contrastive loss L_c (ablation) | LDAM-DRW-SAFA · CIFAR-LT-100 rho=200 | top-1 error: 58.53 -> 57.53 | Improved |
| add head mode seeking loss (ablation) | LDAM-DRW-SAFA · CIFAR-LT-100 rho=200 | top-1 error: 59.89 -> 57.53 | Improved |
| add head mode seeking loss (ablation) | LDAM-DRW-SAFA · CIFAR-LT-100 rho=10 | top-1 error: 41.78 -> 40.89 | Improved |
| add tail mode seeking loss (ablation) | LDAM-DRW-SAFA · CIFAR-LT-100 rho=200 | top-1 error: 58.87 -> 57.53 | Improved |
| add tail mode seeking loss (ablation) | LDAM-DRW-SAFA · CIFAR-LT-100 rho=10 | top-1 error: 42.87 -> 40.89 | Improved |
| add SAFA to LDAM-DRW with ResNet-110 (ablation) | LDAM-DRW · CIFAR-LT-100 rho=200 | top-1 error: 59.13 -> 55.17 | Improved |
| add SAFA to LDAM-DRW with DenseNet-40 (ablation) | LDAM-DRW · CIFAR-LT-100 rho=200 | top-1 error: 58.96 -> 54.87 | Improved |
| add SAFA to LDAM-DRW with ResNet-32 at low imbalance (ablation) | LDAM-DRW · CIFAR-LT-100 rho=10 | top-1 error: 41.22 -> 41.02 | Improved |
| add SAFA to cross-entropy via class-balanced loss | CE loss · CIFAR-LT-100 rho=100 | top-1 error: 61.23 -> 54.13 | Improved |
| replace RSG with SAFA | LDAM-DRS · ImageNet-LT | top-1 accuracy: 51.8 -> 53.1 | Improved |
Closest papers by meaning
- FASA: Feature Augmentation and Sampling Adaptation for Long-Tailed Instance Segmentation (ICCV 2021 · similarity 0.89)
- Feature Space Augmentation for Long-Tailed Data (ECCV 2020 · similarity 0.88)
- Deep Long-Tailed Learning: A Survey (TPAMI 2023 · similarity 0.88)
- Kill Two Birds with One Stone: Rethinking Data Augmentation for Deep Long-tailed Learning (ICLR 2024 · similarity 0.88)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.87)
- Learning from Neighbors: Category Extrapolation for Long-Tail Learning (CVPR 2025 · 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.
