Pothole Counter

A web application that performs video and image analysis, detecting potholes with computer vision.

Web application pages

All recordings view:

Video recording analysis results:

Image recording analysis results:

Map view:

Technical details

Technologies

1. Go - web application and processing management
2. Gin - go-router, handling url paths and templates
3. HTMX, Tailwind - interactive webpages
4. AlpineJS - Google Maps data management
5. Python, Flask - web application for video/image analysis
6. YOLOv8 - video analysis AI model
7. MySQL - data storage
8. Docker - containerizing DB

Structure

Application contains 2 main modules.

ph-manager - webserver on Golang, hosting web application (HTMX) and managing creation and deletion of the recordings.

ph-detector - webserver on Python running a YOLOv8 AI model and processing video/image data to discover and store potholes and find their geolocation using GPX files.

ph-storage - defines the filesystem for this project, storing Videos, Images, Detections, GPX files and Yolo Models.

The main processing flows look like this:

DB

Application uses MySQL database to store data about recordings (uploaded images, videos) and detection (images of the potholes discovered)