As part of the engagement, GlobalLogic applied Artificial Intelligence and Machine Learning (AI/ML) to create several datasets and neural networks to make autonomous cars navigate more safely. We trained classifiers for object detection and understood complex driving environments, learned to navigate through busy city streets with pedestrians, road work, traffic lights.
We leveraged human labellers and automation processes in training the car on navigating through all sorts of situations and environments.
- Video Bounding box: This input enables sensors and software of the car to scan the objects around the vehicle (such as pedestrians, vehicles, cyclists etc)
- Video temporal: This input enables the car to efficiently read the traffic controls, stop signs, crosswalks and other road features to avoid any mishaps
- Image: This enables the software of the car to predict the behaviour of all things around it based on the speed and trajectory, and also helps in choosing other possible paths to avoid risk.