Jiequan Cui, Zhisheng Zhong, Shu Liu, Bei Yu, Jiaya Jia
In this paper, we propose Parametric Contrastive Learning (PaCo) to tackle long-tailed recognition. Based on theoretical analysis, we observe supervised contrastive loss tends to bias on high-frequency classes and thus increases the difficulty of imbalanced learning. We introduce a set of parametric class-wise learnable centers to rebalance from an optimization perspective. Further, we analyze our PaCo loss under a balanced setting. Our analysis demonstrates that PaCo can adaptively enhance the intensity of pushing samples of the same class close as more samples are pulled together with their corresponding centers and benefit hard example learning. Experiments on long-tailed CIFAR, ImageNet, Places, and iNaturalist 2018 manifest the new state-of-the-art for long-tailed recognition. On full ImageNet, models trained with PaCo loss surpass supervised contrastive learning across various ResNet backbones, e.g., our ResNet-200 achieves 81.8% top-1 accuracy. Our code is available at https://github.com/dvlab-research/Parametric-Contrastive-Learning.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Image Classification | Places-LT | Top-1 Accuracy | 41.2 | PaCo |
| Image Classification | CIFAR-10-LT (ρ=10) | Error Rate | 9.14 | PCL |
| Image Classification | ImageNet-LT | Top-1 Accuracy | 60 | PaCo(ResNeXt101-32x4d) |
| Image Classification | ImageNet-LT | Top-1 Accuracy | 58.2 | PaCo(ResNeXt-50) |
| Image Classification | CIFAR-100-LT (ρ=100) | Error Rate | 49.1 | PCL |
| Few-Shot Image Classification | Places-LT | Top-1 Accuracy | 41.2 | PaCo |
| Few-Shot Image Classification | CIFAR-10-LT (ρ=10) | Error Rate | 9.14 | PCL |
| Few-Shot Image Classification | ImageNet-LT | Top-1 Accuracy | 60 | PaCo(ResNeXt101-32x4d) |
| Few-Shot Image Classification | ImageNet-LT | Top-1 Accuracy | 58.2 | PaCo(ResNeXt-50) |
| Few-Shot Image Classification | CIFAR-100-LT (ρ=100) | Error Rate | 49.1 | PCL |
| Generalized Few-Shot Classification | Places-LT | Top-1 Accuracy | 41.2 | PaCo |
| Generalized Few-Shot Classification | CIFAR-10-LT (ρ=10) | Error Rate | 9.14 | PCL |
| Generalized Few-Shot Classification | ImageNet-LT | Top-1 Accuracy | 60 | PaCo(ResNeXt101-32x4d) |
| Generalized Few-Shot Classification | ImageNet-LT | Top-1 Accuracy | 58.2 | PaCo(ResNeXt-50) |
| Generalized Few-Shot Classification | CIFAR-100-LT (ρ=100) | Error Rate | 49.1 | PCL |
| Long-tail Learning | Places-LT | Top-1 Accuracy | 41.2 | PaCo |
| Long-tail Learning | CIFAR-10-LT (ρ=10) | Error Rate | 9.14 | PCL |
| Long-tail Learning | ImageNet-LT | Top-1 Accuracy | 60 | PaCo(ResNeXt101-32x4d) |
| Long-tail Learning | ImageNet-LT | Top-1 Accuracy | 58.2 | PaCo(ResNeXt-50) |
| Long-tail Learning | CIFAR-100-LT (ρ=100) | Error Rate | 49.1 | PCL |
| Generalized Few-Shot Learning | Places-LT | Top-1 Accuracy | 41.2 | PaCo |
| Generalized Few-Shot Learning | CIFAR-10-LT (ρ=10) | Error Rate | 9.14 | PCL |
| Generalized Few-Shot Learning | ImageNet-LT | Top-1 Accuracy | 60 | PaCo(ResNeXt101-32x4d) |
| Generalized Few-Shot Learning | ImageNet-LT | Top-1 Accuracy | 58.2 | PaCo(ResNeXt-50) |
| Generalized Few-Shot Learning | CIFAR-100-LT (ρ=100) | Error Rate | 49.1 | PCL |