shivammaurya14/road-defect-detector ? reverse-engineered prompt

Build me a web app for detecting road potholes and cracks from a live camera or dashcam feed.

I want one clean dashboard where I can switch between three modes: recording raw video with GPS and speed saved to files, running crack segmentation with a heatmap style overlay, and running object detection with boxes, confidence scores, and a live defect counter. It should stay idle until I start it, so it does not use resources when nothing is running.

Also add a map that follows the vehicle and drops pins for detected defects with a small preview when I click them. I need a logs page where I can create manual entries, view snapshots, export CSV, play saved videos, and delete recordings or logs from disk. Please include a simulation mode so I can test everything without a camera or GPS device.

Use Python and Flask for the app, OpenCV for video handling, and make the UI simple and practical for real time use. If anything is unclear, check the current docs online and keep it working end to end.