Abstract: This Scripts enables to track workers in 3-D space with just a RGB Camera Feed and floor plans of the monitored area.
Steps to run:
-
clone the official YoloV7 official repo pip install requirements.txt file
-
Replace the detect and train files in the Yolov7 original repo. Run inference by using trained weights on your frames to identify workers and output their frame-frame feet positions.
-
Run updatedpixel2world, reading in frame-frame feet potions and pick hazard and worksite boundary points from Floorplans. This Outputs individual construction site map frames stored in project images directory.
-
Load project images into VideoGenerator.py to stitch the images into output video. a. Output Video stored as MAPS.mp4