Scene conversion method and system combining instance segmentation and cyclic generative adversarial network

Abstract

The invention provides a scene conversion method and system combining instance segmentation and a cyclic generative adversarial network, and the method comprises: processing a video of a target scene,inputting the processed video into an instance segmentation network, and obtaining segmented scene components, i.e., a mask clipping graph of the target scene; processing targets in the mask clippinggraph of the target scene according to time attribute requirements by using a cyclic generative adversarial network, wherein the generated data is in a state after style migration, and generating a target with unfixed spatial attributes after style migration into a static scene after style migration according to a specific spatial trajectory, so that a scene conversion effect is achieved.

Description

technical field [0001] The present disclosure belongs to the technical field of image recognition, and in particular relates to a scene conversion method and system combining instance segmentation and recurrent generation confrontation network. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] Deep learning is an algorithmic weapon in the era of big data, and it is also a key technology for this round of artificial intelligence explosion. Compared with traditional machine learning algorithms, deep learning technology can continuously improve its performance with the increase of data size, while traditional machine learning algorithms are difficult to use massive data to continuously improve their performance. Convolutional neural network is a deep neural network model widely used in academia and industry. It is widely used in the field of ...

AI Technical Summary

Problems solved by technology

[0005] Therefore, in the practical application of deep learning, such as video surveillance, target detection and other fields, due to the limitation of manpower and material resources, it is not easy to obtain a large amount of high-quality data in these scenarios, and it is difficult to provide data consistent with the testing process in the training process. environment, there may be missing or inconsistency of scene data, such as inconsistency between training scene and test scene, lack of data of a certain season in the same scene, or only day or night data in the same scene

The incompleteness of training and testing scenarios will lead to the loss of scenario data in a specific state in the database, which will affect subsequent data processing and analysis

Images