sekilab/RoadDamageDetector ? reverse-engineered prompt

Build me a simple, easy to follow road damage detection project using this repo’s dataset and notebook as the base.

I want a notebook or lightweight app that helps someone new explore the road damage images, understand the labels, and prepare the data for training an object detection model. It should show sample images with bounding boxes, explain the crack and damage classes in plain language, and walk through how to turn the dataset into something a model can use. If possible, include a basic training and evaluation flow, plus a way to run predictions on a new road image and see the detected damage drawn on top.

Keep the experience beginner friendly, with clear sections, helpful notes, and sensible defaults. Use the files already here, and look up current docs online if you need to. If there’s anything missing, make reasonable assumptions and make it work end to end.