Research map / Learning theory and algorithms / Optimal transport
Counting Trees from Satellite Imagery with Noisy Supervision
Dimitri Gominski, Maurice Mugabowindekwe, Qiue XU, Xiaowei Tong, Martin Brandt, Hieu Le, Rasmus Fensholt, Dimitris Samaras, Loic Landrieu
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 weak ALS-derived labels to training (ablation) | TreeMatch · TinyTrees Rwanda (PlanetScope) | RMSE: 76.1 -> 72.4 | Improved |
| add strong manual labels to training (ablation) | TreeMatch · TinyTrees Rwanda (PlanetScope) | RMSE: 82.4 -> 72.4 | Improved |
| use spatial UOT loss in addition to count loss (ablation) | TreeMatch · TinyTrees Rwanda (PlanetScope) | RMSE: 75.5 -> 72.4 | Improved |
| use unbalanced instead of balanced OT (ablation) | TreeMatch · TinyTrees Rwanda (PlanetScope) | RMSE: 78.3 -> 72.4 | Improved |
| add marginal residual correction of weak labels (ablation) | TreeMatch · TinyTrees Rwanda (PlanetScope) | RMSE: 79.0 -> 72.4 | Improved |
| use curriculum schedule for correction (ablation) | TreeMatch · TinyTrees Rwanda (PlanetScope) | RMSE: 76.3 -> 72.4 | Improved |
| swap ResNet50 for ViT-S backbone (ablation) | TreeMatch · TinyTrees Rwanda (PlanetScope) | RMSE: 72.4 -> 72.5 | No clear effect |
| swap ResNet50 for Swin-S backbone (ablation) | TreeMatch · TinyTrees Rwanda (PlanetScope) | RMSE: 72.4 -> 71.0 | Improved |
| uniform instead of stratified strong/weak sampling (ablation) | TreeMatch · TinyTrees Rwanda (PlanetScope) | RMSE: 72.4 -> 71.4 | Mixed |
| use unbalanced OT instead of DM-Count balanced OT | TreeMatch vs DM-Count · TinyTrees (all three regions) | RMSE: China 65.0 -> 60.6; Rwanda 76.8 -> 72.4; France 158.0 -> 147.2 | Improved |
| add weak supervision to density regression | Density regression · TinyTrees | RMSE | Hurt |
| use pointwise instead of continuous weak labels (ablation) | TreeMatch · TinyTrees China (Gaofen2) | nMAE: 36.6 -> 39.6 | Hurt |
Closest papers by meaning
- OAM-TCD: A globally diverse dataset of high-resolution tree cover maps (NeurIPS 2024 · similarity 0.87)
- ForestLPR: LiDAR Place Recognition in Forests Attentioning Multiple BEV Density Images (CVPR 2025 · similarity 0.86)
- FoMo: Multi-Modal, Multi-Scale and Multi-Task Remote Sensing Foundation Models for Forest Monitoring (AAAI 2025 · similarity 0.86)
- Bringing SAM to new heights: leveraging elevation data for tree crown segmentation from drone imagery (NeurIPS 2025 · similarity 0.86)
- InstructSAM: A Training-free Framework for Instruction-Oriented Remote Sensing Object Recognition (NeurIPS 2025 · similarity 0.85)
- Open-Canopy: Towards Very High Resolution Forest Monitoring (CVPR 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.
