18343291. DISTANCE TO OBSTACLE DETECTION IN AUTONOMOUS MACHINE APPLICATIONS simplified abstract (NVIDIA Corporation)

DISTANCE TO OBSTACLE DETECTION IN AUTONOMOUS MACHINE APPLICATIONS

Organization Name

NVIDIA Corporation

Inventor(s)

Yilin Yang of Santa Clara CA (US)

Bala Siva Sashank Jujjavarapu of Sunnyvale CA (US)

Pekka Janis of Uusimaa (FI)

Zhaoting Ye of Santa Clara CA (US)

Sangmin Oh of San Jose CA (US)

Minwoo Park of Saratoga CA (US)

Daniel Herrera Castro of Uusimaa (FI)

Tommi Koivisto of Uusimaa (FI)

David Nister of Bellevue WA (US)

DISTANCE TO OBSTACLE DETECTION IN AUTONOMOUS MACHINE APPLICATIONS - A simplified explanation of the abstract

This abstract first appeared for US patent application 18343291 titled 'DISTANCE TO OBSTACLE DETECTION IN AUTONOMOUS MACHINE APPLICATIONS

Simplified Explanation: A deep neural network is trained to predict distances to objects using only image data, with the help of ground truth data from depth predicting sensors. Camera adaptation algorithms are used to adjust the network for different camera parameters, and safety bounds are applied to ensure predictions are within a safe range.

Key Features and Innovation:

• Training a deep neural network to predict distances to objects using image data alone
• Utilizing ground truth data from depth predicting sensors to train the network
• Adapting the network for different camera parameters using camera adaptation algorithms
• Applying safety bounds to the predictions to ensure they fall within a safe range

Potential Applications: This technology can be applied in autonomous vehicles, robotics, surveillance systems, and any other systems that require accurate distance prediction using image data.

Problems Solved: This technology addresses the challenge of accurately predicting distances to objects and obstacles using only image data, without the need for additional sensors.

Benefits:

• Improved accuracy in distance prediction using image data
• Adaptability to different camera parameters
• Enhanced safety through the application of safety bounds

Commercial Applications: This technology has potential commercial applications in the automotive industry, robotics industry, security and surveillance sector, and any other industry that requires accurate distance prediction using image data.

Questions about the Technology: 1. How does the deep neural network adapt to different camera parameters? 2. What are the potential limitations of using image data alone for distance prediction?

Frequently Updated Research: There may be ongoing research in the field of computer vision and deep learning related to improving the accuracy and efficiency of distance prediction using image data. Researchers may be exploring new algorithms and techniques to further enhance the capabilities of deep neural networks in this area.

Original Abstract Submitted

In various examples, a deep neural network (DNN) is trained to accurately predict, in deployment, distances to objects and obstacles using image data alone. The DNN may be trained with ground truth data that is generated and encoded using sensor data from any number of depth predicting sensors, such as, without limitation, RADAR sensors, LIDAR sensors, and/or SONAR sensors. Camera adaptation algorithms may be used in various embodiments to adapt the DNN for use with image data generated by cameras with varying parameters—such as varying fields of view. In some examples, a post-processing safety bounds operation may be executed on the predictions of the DNN to ensure that the predictions fall within a safety-permissible range.