Volumetric 3D Gaussian Splatting and precision allometric MRV for smallholder agroforestry. Reconstruct high-density 3D tree models, measure sub-centimeter DBH, and verify stored biomass directly from standard phone video.
Fast and photorealistic 3D volumetric reconstruction your field teams can work with.
Isolate ground plane slope, calculate DBH at 1.3m, and filter stray background particles.
Give your research full compliance with IPCC Tier-1 and Chave biomass equations.
Deterministic global camera pose estimation and multi-view correspondence under a fully permissive commercial BSD-3 license without markers.
Clean-room CUDA kernels from Nerfstudio/UC Berkeley (Apache 2.0) optimize millions of photorealistic 3D Gaussian ellipsoids on Cloud A10G GPUs.
Automated plane extraction isolates forest terrain slope, segments trunk point clouds, and slices cylindrical DBH at 1.30m.
Integrates wood specific gravity (ρ) and geometric DBH for robust pantropical allometric biomass estimations.
Applies Cairns root-to-shoot allometric ratios to ensure complete and auditable carbon accounting.
Instantly export GPS-tagged telemetry, timestamps, and verifiable audit records for carbon credits.
Walk around the tree with any smartphone camera. Smart CV2 extractor isolates sharp, high-overlap frames directly to cloud storage.
Dual-view MASt3R aligns camera poses and generates dense 3D point clouds, followed by fast 3D Gaussian Splatting optimization.
RANSAC detects the ground plane and fits a breast-height cylinder to extract DBH, tree height, carbon stock, and verifiable audit records.