Research map / Label-efficient and robust learning / Recognition and long-tail
Towards Optimal Fine Grained Retrieval via Decorrelated Centralized Loss with Normalize-Scale Layer
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| DGCRL vs CRL-WSL | fine-grained retrieval · CARS196 | Recall@1: 63.9 -> 75.9 | Helped |
| DGCRL vs CRL-WSL | fine-grained retrieval · CUB-200-2011 | Recall@1: 65.9 -> 67.9 | Helped |
| add Gram-Schmidt center decorrelation (lambda 0 -> 0.1) (ablation) | DGCRL · CUB-200-2011 | Recall@1: 66.7 -> 67.9 | Helped |
| increase lambda from 0.1 to 0.4 (ablation) | DGCRL · CUB-200-2011 | Recall@1: 67.9 -> 65.9 | Hurt |
| use DGCRL instead of triplet loss under same N-S layer (ablation) | N-S layer + loss · CUB-200-2011 | Recall@1: 64.4 -> 67.9 | Helped |
| use DGCRL instead of lifted loss (ablation) | N-S layer + loss · CUB-200-2011 | Recall@1: 60.2 -> 67.9 | Helped |
| use DGCRL (softmax form) instead of MSE-based CRL for training (ablation) | N-S layer + loss · CUB-200-2011 | training epochs: 96 -> 20 | Helped |
| increase scale alpha above 16 (ablation) | N-S layer · CUB-200-2011 | Recall@K | Mixed |
Closest papers by meaning
- Embedding Label Structures for Fine-Grained Feature Representation (CVPR 2016 · similarity 0.88)
- Composed Image Retrieval with Text Feedback via Multi-grained Uncertainty Regularization (ICLR 2024 · similarity 0.86)
- NUDGE: Lightweight Non-Parametric Fine-Tuning of Embeddings for Retrieval (ICLR 2025 · similarity 0.85)
- Exploiting Distribution Constraints for Scalable and Efficient Image Retrieval (ICLR 2025 · similarity 0.85)
- Threshold-Consistent Margin Loss for Open-World Deep Metric Learning (ICLR 2024 · similarity 0.85)
- Learning to Align Generative Appearance Priors for Fine-grained Image Retrieval (NeurIPS 2026 · similarity 0.85)
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.