Research map / Detection and segmentation / Open-set and few-shot detection
Estimating Individual Tree Height and Species from UAV Imagery
Jannik Endres, Etienne Laliberté, David Rolnick, Arthur Ouaknine
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 cross-attention task heads over DINOv3 [CLS] linear heads (ablation) | DINOvTree-B · Quebec Trees | delta_1.25: 85.26 -> 88.29 | Improved |
| add cross-attention task heads over DINOv3 [CLS] linear heads (ablation) | DINOvTree-B · Quebec Trees | F1: 83.88 -> 86.89 | Improved |
| add MLP before cross-attention (ablation) | DINOvTree-B · Quebec Trees | F1: 85.00 -> 86.89 | Improved |
| concatenate [CLS] and task token in classification head (ablation) | DINOvTree-B · Quebec Trees | F1: 85.76 -> 86.89 | Improved |
| use ReLU instead of scaled sigmoid for height (ablation) | DINOvTree-B · Quebec Trees | delta_1.25: 87.75 -> 88.29 | Improved |
| use dynamic weight average instead of equal weighting (ablation) | DINOvTree-B · Quebec Trees | delta_1.25: 87.53 -> 88.29 | Improved |
| add PCGrad on top of DWA (ablation) | DINOvTree-B · Quebec Trees | delta_1.25: 88.29 -> 88.09 | No clear effect |
| fine-tune DINOv3-L instead of freezing (ablation) | DINOv3-L · Quebec Trees | delta_1.25: 58.28 -> 88.03 | Improved |
| pretrain on satellite instead of web images (ablation) | DINOv3-L frozen · Quebec Trees | delta_1.25: 58.28 -> 53.07 | Hurt |
| add DSM as fourth input channel (ablation) | ResNet50 · Quebec Trees | delta_1.25: 68.16 -> 68.11 | No clear effect |
| scale DINOvTree from Base to Large | DINOvTree · BCI | delta_1.25: 91.25 -> 94.14 | Improved |
| DINOvTree-L vs DINOv3-L on Quebec Plantations height | DINOvTree-L · Quebec Plantations | delta_1.25: 86.26 -> 85.09 | Hurt |
Closest papers by meaning
- Bringing SAM to new heights: leveraging elevation data for tree crown segmentation from drone imagery (NeurIPS 2025 · similarity 0.90)
- Open-Canopy: Towards Very High Resolution Forest Monitoring (CVPR 2025 · similarity 0.87)
- OAM-TCD: A globally diverse dataset of high-resolution tree cover maps (NeurIPS 2024 · similarity 0.87)
- Estimating Canopy Height at Scale (ICML 2024 · similarity 0.86)
- Capturing Temporal Dynamics in Large-Scale Canopy Tree Height Estimation (ICML 2025 · similarity 0.85)
- FoMo: Multi-Modal, Multi-Scale and Multi-Task Remote Sensing Foundation Models for Forest Monitoring (AAAI 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.
