← Back to Projects
Double Woods Detection System
Computer vision system achieving >95% accuracy for utility infrastructure analysis.
PyTorchComputer VisionCustom Rule Logic
01Pole imagery
02CV detection
03Rule engine
04Maintenance flags
Overview
Deployed CV system for a Fortune 500 utility to detect double wood poles in infrastructure imagery, enabling proactive maintenance planning.
The Problem
Double wood poles (old poles left standing next to their replacements) are a maintenance liability and a compliance issue. Finding them across a Fortune 500 utility's network meant field crews manually reviewing enormous volumes of pole imagery.
The Approach
- 01Trained a PyTorch detection model on utility pole imagery to identify double-wood configurations
- 02Layered a custom rule engine on top of model output to resolve edge cases (occlusion, angles, adjacent structures) that pure detection misses
- 03Tuned the precision/recall trade-off with the client so flagged poles could go straight into maintenance planning without re-review
- 04Integrated results into the client's existing maintenance workflow
The Outcome
>95%detection accuracy in production
Fortune 500utility deployment
Automatedscreening of imagery that previously required manual field review
Key Features
- ✓>95% detection accuracy
- ✓Real-time image processing
- ✓Integration with maintenance systems
- ✓Custom rule engine for edge cases