Unsupervised learning-based motion estimation method

Abstract

The invention discloses an unsupervised learning-based motion estimation method. With the method disclosed by the invention, a problem that the requirement on data is high when a supervised-learning-based deep convolution neural network carries out motion estimation training and especially when true value marking needs to be carried out on lots of training data can be solved. A convolution neural network is trained by using an unsupervised learning method; the requirement on the true value in training data can be reduced by using the method for training and a deep convolution neural network based on a non-general structure is established. In addition, With a calculation model from a rough degree to a precise degree, the trained network model can calculate a motion area with a large motion amplitude.

Description

technical field [0001] The invention relates to a motion estimation method based on unsupervised learning, which belongs to the technical field of computer vision. Background technique [0002] After the AlexNet neural network model achieved unprecedented results in the ImageNet competition, the deep convolutional neural network has received widespread attention, especially in the field of computer vision. The application of convolutional neural networks has solved many problems in the field of computer vision, and also The research field of computer vision has been expanded. But all these progresses and developments benefit from the deep hierarchical structure of the convolutional neural network, as well as a large number of parameters and good training data. [0003] At present, most algorithms for motion estimation rarely use the new technology of deep neural network, because deep neural network is suitable for point-to-point learning, or to find the relationship between...

AI Technical Summary

Problems solved by technology

When it comes to finding differences and correlations between different input data, the performance of deep neural networks becomes less than satisfactory.

FlowNet uses a supervised learning method to train a deep convolutional neural network, but the neural network in FlowNet contains multiple convolutional layers, which makes training a neural network require a large amount of data containing true values

At this stage, there is no standard training database that can provide a large number of ground truth to train deep convolutional neural networks.

Images