Research map / Detection and segmentation / Open-set and few-shot detection
Drone-Based Object Counting by Spatially Regularized Regional Proposal Network
Meng-Ru Hsieh, Yen-Liang Lin, Winston H. Hsu
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 spatial layout regularization (LPN) to the RPN with small anchors (ablation) | RPN+small · PUCPR+ (300 proposals) | average recall: 43.2% -> 49.3% | Improved |
| add spatial layout regularization (LPN) to the RPN with small anchors (ablation) | RPN+small · PUCPR+ (1000 proposals) | average recall: 59.9% -> 62.5% | Improved |
| add spatial layout regularization (LPN) to the RPN with small anchors (ablation) | RPN+small · CARPK (1000 proposals) | average recall: 53.4% -> 57.5% | Improved |
| use smaller anchor boxes on conv4-3 in RPN (ablation) | RPN · PUCPR+ (1000 proposals) | average recall: 21.2% -> 59.9% | Improved |
| use smaller anchor boxes on conv4-3 in RPN (ablation) | RPN · CARPK (1000 proposals) | average recall: 39.2% -> 53.4% | Improved |
| add spatial layout regularization (LPN) in dense scenes versus less dense scenes (ablation) | LPN · PUCPR+ dense vs less dense | relative improvement over RPN-small: 16.30% in dense case vs 8.27% for less dense case | Improved |
| add spatial layout regularization (LPN) at stricter localization (ablation) | RPN+small · PUCPR+ (300 proposals, IoU 0.7) | recall: 54.7% -> 64.4% | Improved |
| use LPN proposals in the counting CNN instead of RPN-small with Faster R-CNN | Faster R-CNN (RPN-small) · PUCPR+ counting | MAE: 39.88 -> 22.76 | Improved |
| use LPN proposals in the counting CNN instead of RPN-small with Faster R-CNN | Faster R-CNN (RPN-small) · CARPK counting | MAE: 24.32 -> 23.80 | Improved |
| fine-tune Faster R-CNN on PUCPR+ instead of using the off-the-shelf PASCAL/COCO model | Faster R-CNN · PUCPR+ counting | MAE: 156.76 -> 111.40 | Improved |
| increase proposals from 400 to 1000 in the counting model (ablation) | Car Counting CNN · PUCPR+ counting | MAE: 22.76 -> 8.04 | Improved |
| use proposal-based counting instead of One-Look Regression | Car Counting CNN · PUCPR+ counting | MAE: 21.88 vs 22.76 | No clear effect |
Closest papers by meaning
- Detection, Tracking, and Counting Meets Drones in Crowds: A Benchmark (CVPR 2021 · similarity 0.89)
- From Open Set to Closed Set: Supervised Spatial Divide-and-Conquer for Object Counting (IJCV 2023 · similarity 0.87)
- Video Individual Counting for Moving Drones (ICCV 2025 · similarity 0.87)
- From Open Set to Closed Set: Counting Objects by Spatial Divide-and-Conquer (ICCV 2019 · similarity 0.86)
- Point Segment and Count: A Generalized Framework for Object Counting (CVPR 2024 · similarity 0.84)
- PointSR: Self-Regularized Point Supervision for Drone-View Object Detection (CVPR 2025 · similarity 0.84)
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.
