tubakhxn/cv_projects — reverse-engineered prompt

Build me a Python based computer vision demo repo with three separate tools I can run on sample videos.

I want one script for autonomous driving footage that can detect cars and pedestrians, track them over time, estimate rough depth from a normal camera view, highlight risky areas, and show a simple driving intelligence view on screen. I also want a power grid inspection tool that can look at drone or inspection footage, find towers and insulators, flag likely faults or problem areas, and show a basic monitoring dashboard. The third tool should analyze football match video by detecting and tracking players, estimating speed and distance covered, showing heatmaps or positioning, and giving simple match insights.

Please keep this clearly educational and research focused, not production ready. Make the code easy to run, with sensible defaults, comments, and sample usage instructions. If anything is missing, add the usual setup files and a simple way to choose which project to run. You can look up current docs online if needed.