Research map / Detection and segmentation / Open-set and few-shot detection
Scaling Up Forest Vision with Synthetic Data
Yihang She, Andrew Blake, David A. Coomes, Srinivasan Keshav
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 |
|---|---|---|---|
| pretrain on synthetic Sim 11,187 before fine-tuning on 3 coniferous plots | A_simp · Coniferous-N | F1 (%): 50.5 -> 79.4 | Improved |
| pretrain on synthetic data before fine-tuning on 2 deciduous plots | A_simp · Deciduous-R | F1 (%): 31.6 -> 70.0 | Improved |
| pretrain on synthetic data before fine-tuning on 2 deciduous plots | A_simp · Deciduous-T | F1 (%): 32.7 -> 59.9 | Improved |
| fine-tune on a single real plot instead of three | A_simp (pretrained) · Coniferous-N | F1 (%): 79.4 -> 77.1 | Hurt |
| reduce real training data from all 42 plots to 3 coniferous plots (from scratch) | A_simp · Coniferous-N | F1 (%): 79.8 -> 50.5 | Hurt |
| add physics-based LiDAR simulation instead of mesh nodal points (ablation) | A_simp pretraining data · Zero-shot Coniferous-C | F1 (%): 5.4 -> 42.6 | Improved |
| add scene diversity (4 to 11 scenes) (ablation) | A_simp pretraining data · Zero-shot Coniferous-C | F1 (%): 42.6 -> 81.5 | Improved |
| add more plots per scene (5 to 17) (ablation) | A_simp pretraining data · Zero-shot Coniferous-C | F1 (%): 81.5 -> 87.0 | Improved |
| add scene diversity (4 to 11 scenes) (ablation) | A_simp pretraining data · Zero-shot Coniferous-S | F1 (%): 52.4 -> 53.0 | No clear effect |
| add more plots per scene (5 to 17) (ablation) | A_simp pretraining data · Zero-shot Deciduous-R | F1 (%): 54.5 -> 61.7 | Improved |
| mix synthetic and real labelled data in pretraining (A_full) | A_full · Deciduous-T | plot-level F1 (%): 69.4 -> 47.4 | Hurt |
| mix synthetic and real labelled data in pretraining (A_full) | A_full · Coniferous-N | plot-level F1 (%): 92.4 -> 92.0 | No clear effect |
Closest papers by meaning
- Bringing SAM to new heights: leveraging elevation data for tree crown segmentation from drone imagery (NeurIPS 2025 · similarity 0.88)
- OAM-TCD: A globally diverse dataset of high-resolution tree cover maps (NeurIPS 2024 · similarity 0.88)
- SVDTree: Semantic Voxel Diffusion for Single Image Tree Reconstruction (CVPR 2024 · similarity 0.85)
- FoMo: Multi-Modal, Multi-Scale and Multi-Task Remote Sensing Foundation Models for Forest Monitoring (AAAI 2025 · similarity 0.85)
- ForestLPR: LiDAR Place Recognition in Forests Attentioning Multiple BEV Density Images (CVPR 2025 · similarity 0.84)
- Reconstructing Animals and the Wild (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.
