This piece of software allows detecting the fact of vehicle examination on video/camera source.
The principle is stupidly easy - to find a vehicle and a person, and to compare the distance between them. If examiner was close enough to the vehicle then examination counts as succeeded.
You can switch debug mode by pressing Ctrl+D.
In this mode, you can see:
- The masks of vehicle anmd examiner
- Their bounding boxes
- Circles in which they find their position on a previous frame
- Expanded bounding boxes that are used to decide the fact of examination.
- Setup required modules from
requirements.txt - Download weights file to
./data/folder - Run
main.py
In this software we use a modified version of Mask-RCNN. We have removed all the parts that needs for model training. Also, the big part of utilitary functions has been removed as it's not needed. Some functions had been moved or refactored to increase development and runtime efficiency.
This software is provided as-is and gives no warranties at all.