manasvi-glitch/NHAI---AI-assisted-Reflectivity-Assessment ? reverse-engineered prompt

Build me a simple Streamlit app that lets a user upload a road image or a short video clip and then estimates how reflective the road markings look.

I want the app to show a clear score from 0 to 100, plus an easy label like good, moderate, or poor. If the score is low, it should also warn that the marking may need maintenance. Please highlight the reflective parts of the image in yellow and the detected edges in orange so the result is easy to understand at a glance.

Keep it lightweight and local, using basic image processing only, no machine learning training. The flow should be straightforward, take the uploaded image or a frame from the video, process it, and show the result in the browser. Make it look clean and usable, and include any small details needed so I can run it with one command. If you need to, look up current Streamlit or OpenCV docs online.