UNSTRUCTURED
The solutions proposed so far to the problem of recognising a chess diagram from an image containing a standard chess set are not usable in a real scenario, because the constant mistakes in recognising different chess pieces, such as pawns versus bishops, for example, make the whole recognition process useless. There is a great need to combine the computing power of chess engines with a physical chessboard, especially during individual chess training. Therefore, this article proposes a method for identifying chess positions on a two-dimensional chess set, which allows obtaining close to perfect accuracy during the classification of chess pieces. The generation of a custom data set using the specific Magne 2D chess set, which was designed in 2019 and is the first of its kind in the world, allowed the proposed model to have the aforementioned accuracy. Furthermore, the recognition process is supported by a cross-platform mobile application, which includes the necessary artificial intelligence features to recognize a chess diagram.