View source for 18367888. UNSUPERVISED DEPTH PREDICTION NEURAL NETWORKS simplified abstract (GOOGLE LLC)

=UNSUPERVISED DEPTH PREDICTION NEURAL NETWORKS=


==Organization Name==

[[:Category:GOOGLE LLC|GOOGLE LLC]]


[[Category:GOOGLE LLC]]

==Inventor(s)==

[[:Category:Vincent Michael Casser of Cambridge MA (US)|Vincent Michael Casser of Cambridge MA (US)]][[Category:Vincent Michael Casser of Cambridge MA (US)]]

[[:Category:Soeren Pirk of Palo Alto CA (US)|Soeren Pirk of Palo Alto CA (US)]][[Category:Soeren Pirk of Palo Alto CA (US)]]

[[:Category:Reza Mahjourian of Austin TX (US)|Reza Mahjourian of Austin TX (US)]][[Category:Reza Mahjourian of Austin TX (US)]]

[[:Category:Anelia Angelova of Sunnyvale CA (US)|Anelia Angelova of Sunnyvale CA (US)]][[Category:Anelia Angelova of Sunnyvale CA (US)]]

==UNSUPERVISED DEPTH PREDICTION NEURAL NETWORKS - A simplified explanation of the abstract==


This abstract first appeared for US patent application 18367888 titled 'UNSUPERVISED DEPTH PREDICTION NEURAL NETWORKS

==Simplified Explanation==

The patent application describes a system for generating a depth output for an image. Here are the key points:

* The system takes input images of the same scene, each with potential objects.
* It generates a background image for each input image and analyzes them to determine camera motion.
* For each potential object, it generates an object motion output based on the input images and camera motion.
* The system uses a depth prediction neural network to process a particular input image and generate a depth output.
* The parameters of the depth prediction neural network are updated based on the depth output, camera motion, and object motion.

Potential applications of this technology:

* Computer vision systems that require accurate depth information for image processing.
* Augmented reality applications that need to understand the depth of objects in the real world.
* Autonomous vehicles that rely on depth perception for navigation and obstacle avoidance.

Problems solved by this technology:

* Accurately determining the depth of objects in an image can be challenging, especially when there is camera motion and potential objects in the scene.
* This system addresses these challenges by considering camera motion and object motion to generate a more accurate depth output.

Benefits of this technology:

* Improved accuracy in generating depth outputs for images.
* The ability to update the parameters of the depth prediction neural network based on the specific scene and camera/object motion, leading to better performance over time.
* The system can be applied to various applications that require depth information, enhancing their capabilities.


==Original Abstract Submitted==

A system for generating a depth output for an image is described. The system receives input images that depict the same scene, each input image including one or more potential objects. The system generates, for each input image, a respective background image and processes the background images to generate a camera motion output that characterizes the motion of the camera between the input images. For each potential object, the system generates a respective object motion output for the potential object based on the input images and the camera motion output. The system processes a particular input image of the input images using a depth prediction neural network (NN) to generate a depth output for the particular input image, and updates the current values of parameters of the depth prediction NN based on the particular depth output, the camera motion output, and the object motion outputs for the potential objects.



[[Category:G06T7/55]]
[[Category:G06T7/246]]
[[Category:G06N3/088]]
[[Category:G06T3/00]]
[[Category:G06N3/045]]