Guardian
A system designed to minimize the risks of injury or fatality during operations performed by workers on railroad infrastructure facilities.
Client
TechTrans
Service
System for monitoring the safety of workers on the railroads
Date
May 5, 2021
Challenge
The complex shall be able to manually and automatically determine the guard zone at the objects of railroad infrastructure, guaranteed detection of a person in the specified guard zone and automatic determination of man crossing the boundaries of the guard zone with the issuance of appropriate warning signals.
Solution
The Guardian contained:
- OpenCV - to work with computer vision algorithms and cameras
- PyTorch - framerate for neural networks
- TensorRT - framework for neural network acceleration (selective)
- NumPy - a framework for working with arrays
- Shapely - for polygon shaping
- ROS2 Galactic (selective) - to provide interconnection between the nodes of the system
- clearML - for training neural networks, hypothesis testing and metrics estimation
- RoboFlow, CVAT - for annotation, dataset augmentation
More details can be seen here
