Ashutosh27ind/Vehicle-detection-tracking-and-classification — reverse-engineered prompt

Build me a simple notebook based project that can take a road traffic video, detect moving vehicles frame by frame, track them as they pass, and count each one when it crosses a virtual line on the lane. After a vehicle crosses the line, crop it from the frame and run a CNN classifier so I can label it by type, for example two wheeler, four wheeler, truck, auto rickshaw, bicycle, or whatever classes make sense from the data we have.

I want this to feel like a complete demo I can run and understand, not just scattered experiments. Please clean up the workflow into a few clear notebooks or scripts, include preprocessing, detection, tracking, classification, and show the counted results on the video output. If a trained model or sample video is needed, wire that in and explain how to use it. Add simple setup and run instructions so I can test it on my own traffic clip too. Look up current docs online if you need to.