Assigning obstacles to lanes using neural networks for autonomous machine applications

This invention describes a computer system for autonomous machines, such as self-driving cars, to understand their environment. It uses a specialized computer program, called a neural network, to analyze real-time sensor data. This program identifies objects and simultaneously determines which lane each object occupies by generating detailed "masks" that indicate both the object type and its lane for every pixel in the sensor data, helping the machine make decisions about its surroundings.