Zihan Zhang, Xiang Xiang
The real-world data distribution is essentially long-tailed, which poses great challenge to the deep model. In this work, we propose a new method, Gradual Balanced Loss and Adaptive Feature Generator (GLAG) to alleviate imbalance. GLAG first learns a balanced and robust feature model with Gradual Balanced Loss, then fixes the feature model and augments the under-represented tail classes on the feature level with the knowledge from well-represented head classes. And the generated samples are mixed up with real training samples during training epochs. Gradual Balanced Loss is a general loss and it can combine with different decoupled training methods to improve the original performance. State-of-the-art results have been achieved on long-tail datasets such as CIFAR100-LT, ImageNetLT, and iNaturalist, which demonstrates the effectiveness of GLAG for long-tailed visual recognition.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Image Classification | CIFAR-100-LT (ρ=10) | Error Rate | 35.5 | GLAG |
| Image Classification | CIFAR-100-LT (ρ=100) | Error Rate | 48.3 | GLAG |
| Few-Shot Image Classification | CIFAR-100-LT (ρ=10) | Error Rate | 35.5 | GLAG |
| Few-Shot Image Classification | CIFAR-100-LT (ρ=100) | Error Rate | 48.3 | GLAG |
| Generalized Few-Shot Classification | CIFAR-100-LT (ρ=10) | Error Rate | 35.5 | GLAG |
| Generalized Few-Shot Classification | CIFAR-100-LT (ρ=100) | Error Rate | 48.3 | GLAG |
| Long-tail Learning | CIFAR-100-LT (ρ=10) | Error Rate | 35.5 | GLAG |
| Long-tail Learning | CIFAR-100-LT (ρ=100) | Error Rate | 48.3 | GLAG |
| Generalized Few-Shot Learning | CIFAR-100-LT (ρ=10) | Error Rate | 35.5 | GLAG |
| Generalized Few-Shot Learning | CIFAR-100-LT (ρ=100) | Error Rate | 48.3 | GLAG |