Research map / Detection and segmentation / Open-set and few-shot detection
Balanced Hierarchical Contrastive Learning with Decoupled Queries for Fine-grained Object Detection in Remote Sensing Images
Jingzhou Chen, Dexin Chen, Fengchao Xiong, Yuntao Qian, Liang Xiao
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 |
|---|---|---|---|
| add query decoupling (classification vs localization queries) (ablation) | RHINO · ShipRSImageNet val | AP50:95: 59.78 -> 60.99 | Improved |
| add query decoupling (ablation) | OrientedFormer · ShipRSImageNet val | AP50:95: 63.17 -> 63.60 | Improved |
| add HCL on top of decoupling (ablation) | OrientedFormer · ShipRSImageNet val | AP50:95: 63.60 -> 64.12 | Improved |
| add HCL on top of decoupling (ablation) | OrientedFormer · FAIR1M-v1.0 test | AP50: 41.38 -> 41.14 | Hurt |
| use balanced HCL instead of HCL (ablation) | OrientedFormer · FAIR1M-v1.0 test | AP50: 41.14 -> 41.66 | Improved |
| use balanced HCL instead of HCL (ablation) | OrientedFormer · ShipRSImageNet val | AP50:95: 64.12 -> 64.32 | Improved |
| use balanced HCL instead of HCL (ablation) | RHINO · ShipRSImageNet val | AP50:95: 61.24 -> 61.41 | Improved |
| add EMA class prototypes to BHCL (ablation) | OrientedFormer · ShipRSImageNet val | AP50:95: 63.34 -> 64.32 | Improved |
| use EMA prototypes instead of classifier weights (ablation) | OrientedFormer · ShipRSImageNet val | AP50:95: 64.26 -> 64.32 | No clear effect |
| use EMA update instead of gradient-based prototype update (ablation) | OrientedFormer · ShipRSImageNet val | AP50:95: 63.88 -> 64.32 | Improved |
| use full-epoch averaging instead of EMA (ablation) | OrientedFormer · ShipRSImageNet val | AP50:95: 64.32 -> 64.37 | No clear effect |
| use proposed method versus OrientedFormer | Ours · ShipRSImageNet val | AP50:95: 63.2 -> 64.3 | Improved |
Closest papers by meaning
- SLIP-RS: Structured-Attribute Language-Image Pre-Training for Remote Sensing Object Detection (ICML 2026 · similarity 0.87)
- Geography-Aware Self-Supervised Learning (ICCV 2021 · similarity 0.87)
- Sparse and Complete Latent Organization for Geospatial Semantic Segmentation (CVPR 2022 · similarity 0.87)
- Long-tailed Object Detection Pretraining: Dynamic Rebalancing Contrastive Learning with Dual Reconstruction (NeurIPS 2024 · similarity 0.86)
- InstructSAM: A Training-free Framework for Instruction-Oriented Remote Sensing Object Recognition (NeurIPS 2025 · similarity 0.86)
- In2NeCT: Inter-class and Intra-class Neural Collapse Tuning for Semantic Segmentation of Imbalanced Remote Sensing Images (AAAI 2025 · 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.
