Research map / Label-efficient and robust learning / Semi-supervised learning
Probabilistic Contrastive Learning for Long-Tailed Visual Recognition
Chaoqun Du, Yulin Wang, Shiji Song, Gao Ming Huang
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 |
|---|---|---|---|
| model class features with vMF distribution instead of EMA prototype (ablation) | ProCo · CIFAR-100-LT, imbalance 100 (200 epochs) | top-1 accuracy: 51.6 -> 52.8 | Improved |
| model class features with vMF instead of normal distribution (ablation) | ProCo · CIFAR-100-LT, imbalance 100 (200 epochs) | top-1 accuracy: 52.1 -> 52.8 | Improved |
| use vMF instead of centroid prototype (ablation) | ProCo · CIFAR-100-LT, imbalance 100 (200 epochs) | top-1 accuracy: 52.0 -> 52.8 | Improved |
| add supervised contrastive loss to ProCo (ablation) | ProCo · CIFAR-100-LT, imbalance 100 (200 epochs) | top-1 accuracy: 52.8 -> 52.6 | Hurt |
| add logit adjustment to learnable class representation (ablation) | ProCo · CIFAR-100-LT, imbalance 100 (200 epochs) | top-1 accuracy: 50.5 -> 50.6 | No clear effect |
| use L_in instead of L_out (ablation) | ProCo · CIFAR-100-LT, imbalance 100 | top-1 accuracy: 52.0 -> 52.8 | Improved |
| use L_in instead of L_out (ablation) | ProCo · CIFAR-100-LT, imbalance 10 | top-1 accuracy: 65.1 -> 65.5 | Improved |
| add AutoAugment (ablation) | ProCo · CIFAR-100-LT, imbalance 100 (200 epochs) | top-1 accuracy: 50.5 -> 52.8 | Improved |
| add AutoAugment (ablation) | BCL · CIFAR-100-LT, imbalance 100 (200 epochs) | top-1 accuracy: 50.1 -> 51.9 | Improved |
| add ProCo to Faster R-CNN | Faster R-CNN · LVIS v1 | AP_b: 22.1 -> 24.7 | Improved |
| add ProCo to Mask R-CNN | Mask R-CNN · LVIS v1 | AP_m: 21.7 -> 24.7 | Improved |
| add ProCo to NCL ensemble | NCL · ImageNet-LT (180 epochs) | top-1 accuracy: 59.2 -> 60.2 | Improved |
Closest papers by meaning
- Continuous Contrastive Learning for Long-Tailed Semi-Supervised Recognition (NeurIPS 2024 · similarity 0.90)
- Decoupled Contrastive Learning for Long-Tailed Recognition (AAAI 2024 · similarity 0.89)
- Long-tailed Object Detection Pretraining: Dynamic Rebalancing Contrastive Learning with Dual Reconstruction (NeurIPS 2024 · similarity 0.89)
- A Tale of Two Classes: Adapting Supervised Contrastive Learning to Binary Imbalanced Datasets (CVPR 2025 · similarity 0.87)
- Robust Visual Recognition with Class-Imbalanced Open-World Noisy Data (AAAI 2024 · similarity 0.86)
- BaCon: Boosting Imbalanced Semi-supervised Learning via Balanced Feature-Level Contrastive Learning (AAAI 2024 · 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.
