Rahul Vigneswaran, Marc T. Law, Vineeth N. Balasubramanian, Makarand Tapaswi
The visual world naturally exhibits an imbalance in the number of object or scene instances resulting in a \emph{long-tailed distribution}. This imbalance poses significant challenges for classification models based on deep learning. Oversampling instances of the tail classes attempts to solve this imbalance. However, the limited visual diversity results in a network with poor representation ability. A simple counter to this is decoupling the representation and classifier networks and using oversampling only to train the classifier. In this paper, instead of repeatedly re-sampling the same image (and thereby features), we explore a direction that attempts to generate meaningful features by estimating the tail category's distribution. Inspired by ideas from recent work on few-shot learning, we create calibrated distributions to sample additional features that are subsequently used to train the classifier. Through several experiments on the CIFAR-100-LT (long-tail) dataset with varying imbalance factors and on mini-ImageNet-LT (long-tail), we show the efficacy of our approach and establish a new state-of-the-art. We also present a qualitative analysis of generated features using t-SNE visualizations and analyze the nearest neighbors used to calibrate the tail class distributions. Our code is available at https://github.com/rahulvigneswaran/TailCalibX.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Image Classification | CIFAR-100-LT (ρ=50) | Error Rate | 49.1 | CBD+TailCalibX |
| Image Classification | CIFAR-100-LT (ρ=10) | Error Rate | 38.87 | CBD+TailCalibX |
| Image Classification | CIFAR-100-LT (ρ=100) | Error Rate | 53.41 | CBD+TailCalibX |
| Image Classification | mini-ImageNet-LT | Error Rate | 55.27 | TailCalibX |
| Few-Shot Image Classification | CIFAR-100-LT (ρ=50) | Error Rate | 49.1 | CBD+TailCalibX |
| Few-Shot Image Classification | CIFAR-100-LT (ρ=10) | Error Rate | 38.87 | CBD+TailCalibX |
| Few-Shot Image Classification | CIFAR-100-LT (ρ=100) | Error Rate | 53.41 | CBD+TailCalibX |
| Few-Shot Image Classification | mini-ImageNet-LT | Error Rate | 55.27 | TailCalibX |
| Generalized Few-Shot Classification | CIFAR-100-LT (ρ=50) | Error Rate | 49.1 | CBD+TailCalibX |
| Generalized Few-Shot Classification | CIFAR-100-LT (ρ=10) | Error Rate | 38.87 | CBD+TailCalibX |
| Generalized Few-Shot Classification | CIFAR-100-LT (ρ=100) | Error Rate | 53.41 | CBD+TailCalibX |
| Generalized Few-Shot Classification | mini-ImageNet-LT | Error Rate | 55.27 | TailCalibX |
| Long-tail Learning | CIFAR-100-LT (ρ=50) | Error Rate | 49.1 | CBD+TailCalibX |
| Long-tail Learning | CIFAR-100-LT (ρ=10) | Error Rate | 38.87 | CBD+TailCalibX |
| Long-tail Learning | CIFAR-100-LT (ρ=100) | Error Rate | 53.41 | CBD+TailCalibX |
| Long-tail Learning | mini-ImageNet-LT | Error Rate | 55.27 | TailCalibX |
| Generalized Few-Shot Learning | CIFAR-100-LT (ρ=50) | Error Rate | 49.1 | CBD+TailCalibX |
| Generalized Few-Shot Learning | CIFAR-100-LT (ρ=10) | Error Rate | 38.87 | CBD+TailCalibX |
| Generalized Few-Shot Learning | CIFAR-100-LT (ρ=100) | Error Rate | 53.41 | CBD+TailCalibX |
| Generalized Few-Shot Learning | mini-ImageNet-LT | Error Rate | 55.27 | TailCalibX |