Optical flow determination system

Abstract

A generative adversarial network (GAN) system includes a generator sub-network configured to examine images of an object moving relative to a viewer of the object. The generator sub-network also is configured to generate one or more distribution-based images based on the images that were examined. The system also includes a discriminator sub-network configured to examine the one or more distribution-based images to determine whether the one or more distribution-based images accurately represent the object. A predicted optical flow of the object is represented by relative movement of the objectas shown in the one or more distribution-based images.

Description

technical field [0001] This application relates to image analysis systems using one or more neural networks. Background technique [0002] Neural networks can be used to analyze images for a variety of purposes. For example, some neural networks examine images to identify objects depicted in the images. Some sets of images often represent or can represent frames of video showing relative motion between objects depicted in the images and the viewer of the images. Optical flow describes this relative motion of objects in the visible field with respect to the observer. Computing optical flow accurately and efficiently can be used in various computer vision applications such as object detection, tracking, motion detection, robot navigation, three-dimensional (3D) reconstruction and segmentation, and many more. [0003] However, calculating optical flow accurately and efficiently can be difficult to achieve. This computation is equivalent to finding pixel correspondences betw...

AI Technical Summary

Problems solved by technology

[0003] However, calculating optical flow accurately and efficiently can be difficult to achieve

This calculation is equivalent to finding the pixel correspondence between consecutive frames, which is a time-consuming task

Sometimes, there is not any exact pixel intensity correspondence, or the solution is not unique, which frustrates or hinders the determination of optical flow

Images