casedone/people-counting — reverse-engineered prompt
Reverse engineered prompt
Build me a Python app that counts people in videos using YOLO models. I want to upload or point to a video, draw a counting line on a frame, then process the video and see how many people crossed the line in each direction.
Please include a simple web interface with Gradio so a non technical user can pick a video, choose a YOLO model type like YOLOv8, YOLOv10, or YOLOv11, set confidence, choose whether to save the annotated video, and see the final counts. The output video should show detected people, tracking ids, the counting line, and the live count overlay.
Also include command line scripts for processing one video, selecting the line interactively, cleaning input and output folders, and batch processing multiple videos from a JSON or CSV job file. Add a setup demo script that installs what is needed, downloads a small model and sample video, then runs a demo. Use local ffmpeg and ffprobe binaries from a bin folder if needed. Look up current docs online if you need to.
Want more depth? Deep Reverse