Chess situation conversion model establishment and conversion method and device based on representation learning

Abstract

The invention relates to a representation learning-based chess situation conversion model establishment method and device, and a representation learning-based chess situation conversion method and device, and belongs to the technical field of image recognition. The method comprises the following steps: extracting features of a pre-marked chess game picture through a backbone network, and extracting the features in a layered manner to generate a multi-scale feature layer; inputting the multi-scale feature layer into a target detection special layer to generate a plurality of candidate frames, and screening the plurality of candidate frames to obtain a standard candidate frame; calculating a loss value between the standard candidate box and a pre-annotation, adjusting model parameters through gradient descent to enable the loss value not to be reduced or to reach a preset number of iterations, and obtaining a chess situation conversion model; through the method and the device, the problems in the prior art that only existing templates in a template library can be matched, accurate matching cannot be performed on a new image, a template matching method is strict in image size requirement, and if the image sizes are not matched, the matching precision is also reduced are solved.

Description

technical field [0001] The present application belongs to the technical field of image recognition, and in particular relates to a method and device for establishing and transforming a chess situation transformation model based on representation learning. Background technique [0002] Existing chess exercises, you can only find chess game questions on the Internet, and you cannot directly obtain the FEN value that the computer can understand. Generally, you need to manually place the chess pieces, which takes a lot of time and is inefficient. At the same time, it is not conducive to using a computer to save the chess game situation. , which is convenient for subsequent analysis of the chess game, usually only pictures can be saved. When it needs to be used, it also needs to be manually restored in the chessboard according to the saved chess position pictures, which is not conducive to the teaching and learning of the chess game; [0003] In the existing patent technology, th...

AI Technical Summary

Problems solved by technology

[0004] For this reason, this application provides a chess game situation transformation model establishment and transformation method and device based on characterization learning, which is used to solve the problem that in the prior art, only existing templates in the template library can be matched, and accurate matching cannot be performed for new images. , and the method based on template matching has strict requirements on the image size. If the image size does not match, the matching accuracy will also be reduced.

Images