Ajd

Reporting a pothole is a hassle, so most never get reported. RoadWatch fixes that: upload a photo and Muse Spark’s vision model locates each pothole and estimates its depth, size, and nearby hazards such as traffic lane or standing water. A transparent scoring function turns those findings into a 0-100 severity score and a priority level (Low to Critical), so the result is explainable and consistent. Muse Spark then writes a polite, factual complaint with the location, a map link, and a requested repair timeframe, which the user can edit, download, or email. The model handles perception and writing, and plain code produces the priority score. Built with Python, Streamlit, and the Meta Model API. Open source under the MIT license.