AI Solutions
Tree-Quality Computer Vision
Field-deployable computer-vision system that grades timber quality from smartphone photos.
About this project
Forestry buyers needed faster, more consistent grading of standing timber than expert visual inspection allowed.
Solution
Trained a custom vision model on annotated bark/canopy imagery, packaged into a mobile-friendly inference endpoint with offline fallback.
Technology
- PyTorch
- ONNX Runtime
- Swift
- Kotlin
- FastAPI
- AWS Lambda
Impact
Reduced grading time per stand from 2 hours to under 15 minutes; inter-rater agreement improved from 64% to 91%.