Case Study
The City of Houston needed to automate the process of analyzing pavement images to classify roads and calculate measurements like width and area. Manually performing these tasks was time-consuming and tedious.
Situational Analysis
- Road maintenance teams took thousands of pavement images annually
- Images were manually reviewed to classify road types and take measurements
- This process was slow, expensive, and prone to human error
- Needed to automate classification and measurement to save time
Solution
- Built a convolutional neural network (CNN) model using TensorFlow
- Trained model on sample pavement images with classification labels
- Deployed model to classify new images in real-time
- Added algorithms to calculate width, area from classified images
- Developed a web app for road crews to upload images
Technology
- TensorFlow open source library for machine learning
- Python for model training and deployment
- ReactJS web framework for image upload app
- Cloud hosting for scalable deployment
Business Impact
- Reduced road measurement workload by 75%
- Significantly decreased time and costs
- Enabled road teams to focus on higher value work
- Provided scalable solution to grow with road network