snehilsanyal/Construction-Site-Safety-PPE-Detection — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a construction site safety detection project that can look at photos or videos of workers and mark what safety gear is present or missing. I want it to detect people, hardhats, masks, safety vests, safety cones, machinery, vehicles, and also flag workers with no hardhat, no mask, or no safety vest.

Use YOLOv8 with the construction site safety dataset from Roboflow or Kaggle if needed. Train a small custom model, save the best model, and include simple examples showing predictions on test images and sample videos. I’d like the results to include labelled bounding boxes, a confusion matrix, training charts, and a few output images so it’s easy to see if it worked.

Please organize it so someone can run the notebook or script again, with clear setup steps and folders for data, models, results, and outputs. Look up current YOLOv8 docs online if you need to.

Want more depth? Deep Reverse