Research map / Detection and segmentation / CNN features
Dynamic Refinement Network for Oriented and Densely Packed Object Detection
Check 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 |
|---|---|---|---|
| Add the full DRN to CenterNet | CenterNet · DOTA test (OBB) | mAP: 65.04 -> 70.70 | Helped |
| Add the full DRN to CenterNet | CenterNet · HRSC2016 test (COCO evaluation) | mAP: 63.5 -> 65.6 | Helped |
| Add the full DRN to CenterNet | CenterNet · SKU110K-R test | mAP: 54.4 -> 55.9 | Helped |
| Add the full DRN to CenterNet | CenterNet · SKU110K test | mAP: 55.8 -> 56.9 | Helped |
| Use three receptive-field branches (3×3, 1×3, and 3×1) with rotation convolution (ablation) | FSM on CenterNet · DOTA validation | AP50: 63.4 -> 64.4 | Helped |
| Use rotation convolution rather than DCN in the three-branch FSM (ablation) | FSM on CenterNet · DOTA validation | AP50: 63.9 -> 64.4 | Helped |
| Add a 1×3 branch to the one-branch rotation-convolution FSM (ablation) | FSM on CenterNet · DOTA validation | AP50: 63.9 -> 64.2 | Helped |
| Add a 3×1 branch to the two-branch rotation-convolution FSM (ablation) | FSM on CenterNet · DOTA validation | AP50: 64.2 -> 64.4 | Helped |
| Add dynamic refinement to the classification head (ablation) | CenterNet · DOTA validation | AP50: 63.4 -> 64.1 | Helped |
| Add dynamic refinement to scale regression (ablation) | CenterNet · DOTA validation | AP50: 63.4 -> 64.1 | Helped |
| Add dynamic refinement to angle regression (ablation) | CenterNet · DOTA validation | AP50: 63.4 -> 63.5 | No clear effect |
| Add dynamic refinement to offset regression (ablation) | CenterNet · DOTA validation | AP50: 63.4 -> 63.4 | No clear effect |
Closest papers by meaning
- DynamicDet: A Unified Dynamic Architecture for Object Detection (CVPR 2023 · similarity 0.89)
- ObjectNet: A large-scale bias-controlled dataset for pushing the limits of object recognition models (NeurIPS 2019 · similarity 0.88)
- Spatially Invariant Unsupervised Object Detection with Convolutional Neural Networks (AAAI 2019 · similarity 0.88)
- Domain Adaptive Object Detection via Dynamic Causal Refinement (ICML 2026 · similarity 0.87)
- Single-Shot Refinement Neural Network for Object Detection (CVPR 2018 · similarity 0.87)
- Straight to Shapes: Real-Time Detection of Encoded Shapes (CVPR 2017 · similarity 0.87)
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.