JianFeng Wang, Thomas Lukasiewicz, Xiaolin Hu, Jianfei Cai, Zhenghua Xu
Imbalanced datasets widely exist in practice and area great challenge for training deep neural models with agood generalization on infrequent classes. In this work, wepropose a new rare-class sample generator (RSG) to solvethis problem. RSG aims to generate some new samplesfor rare classes during training, and it has in particularthe following advantages: (1) it is convenient to use andhighly versatile, because it can be easily integrated intoany kind of convolutional neural network, and it works wellwhen combined with different loss functions, and (2) it isonly used during the training phase, and therefore, no ad-ditional burden is imposed on deep neural networks duringthe testing phase. In extensive experimental evaluations, weverify the effectiveness of RSG. Furthermore, by leveragingRSG, we obtain competitive results on Imbalanced CIFARand new state-of-the-art results on Places-LT, ImageNet-LT, and iNaturalist 2018. The source code is available at https://github.com/Jianf-Wang/RSG.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Image Classification | Places-LT | Top-1 Accuracy | 39.3 | LDAM-DRS-RSG |
| Image Classification | CIFAR-100-LT (ρ=50) | Error Rate | 51.5 | LDAM-DRW-RSG |
| Image Classification | ImageNet-LT | Top-1 Accuracy | 51.8 | LDAM-DRS-RSG |
| Image Classification | CIFAR-100-LT (ρ=100) | Error Rate | 55.5 | LDAM-DRW-RSG |
| Few-Shot Image Classification | Places-LT | Top-1 Accuracy | 39.3 | LDAM-DRS-RSG |
| Few-Shot Image Classification | CIFAR-100-LT (ρ=50) | Error Rate | 51.5 | LDAM-DRW-RSG |
| Few-Shot Image Classification | ImageNet-LT | Top-1 Accuracy | 51.8 | LDAM-DRS-RSG |
| Few-Shot Image Classification | CIFAR-100-LT (ρ=100) | Error Rate | 55.5 | LDAM-DRW-RSG |
| Generalized Few-Shot Classification | Places-LT | Top-1 Accuracy | 39.3 | LDAM-DRS-RSG |
| Generalized Few-Shot Classification | CIFAR-100-LT (ρ=50) | Error Rate | 51.5 | LDAM-DRW-RSG |
| Generalized Few-Shot Classification | ImageNet-LT | Top-1 Accuracy | 51.8 | LDAM-DRS-RSG |
| Generalized Few-Shot Classification | CIFAR-100-LT (ρ=100) | Error Rate | 55.5 | LDAM-DRW-RSG |
| Long-tail Learning | Places-LT | Top-1 Accuracy | 39.3 | LDAM-DRS-RSG |
| Long-tail Learning | CIFAR-100-LT (ρ=50) | Error Rate | 51.5 | LDAM-DRW-RSG |
| Long-tail Learning | ImageNet-LT | Top-1 Accuracy | 51.8 | LDAM-DRS-RSG |
| Long-tail Learning | CIFAR-100-LT (ρ=100) | Error Rate | 55.5 | LDAM-DRW-RSG |
| Generalized Few-Shot Learning | Places-LT | Top-1 Accuracy | 39.3 | LDAM-DRS-RSG |
| Generalized Few-Shot Learning | CIFAR-100-LT (ρ=50) | Error Rate | 51.5 | LDAM-DRW-RSG |
| Generalized Few-Shot Learning | ImageNet-LT | Top-1 Accuracy | 51.8 | LDAM-DRS-RSG |
| Generalized Few-Shot Learning | CIFAR-100-LT (ρ=100) | Error Rate | 55.5 | LDAM-DRW-RSG |