NVIDIA’s Neural Networks Detect Distance Between Vehicles and Pedestrians in 3D

Nvidia Neural Networks

The first basic ability you need to develop before driving a vehicle on the road is to judge the distance between you and the adjacent vehicle or object you’re coming across. However, we often tend to lose attention on the road when someone interrupts us or even a phone call can do the trick. To overcome this problem, researchers at NVIDIA have come up with a solution that makes use of neural networks.

NVIDIA extracted data from a single camera to perform distance-to-object detection by utilizing deep neural networks. You might be wondering why the brand didn’t make use of dual cameras seen on most modern cars for better efficiency. Well, NVIDIA has a solid reason for that. Consider a self-driving car with a dual-camera stereo vision system. While dual cameras will help to proactively judge various parameters, it may lead to a condition called “timing misalignment” which occurs if either of the cameras goes out of sync. As a result, inaccurate results may be derived that causes critical damage to life and property in sensitive situations.

The brand uses convolutional neural networks that depend on radar and lidar sensor data as a base for performing the estimations. This would enable the distance calculation to be performed irrespective of the terrain of the road so you can rely on the values predicted by the neural network.

The green boxes in the above image represent the object detection and the rapidly changing numbers on top of the boxes denote the distance between your vehicle and the detected object. Object detection is not only limited to vehicles. It can also detect pedestrians on the road so that accident situations can be avoided.

The vision of NVIDIA is to make use of this 3D distance calculation methodology for powering tasks like autonomous cruise control and automated lane changing. We’ll have to wait until the technology matures to provide completely reliable results to be used on a daily basis.

comment Comments 0
Leave a Reply