The problem of data classification realized by the convolutional networks, although successfully implemented using classical methods, is also important in the area of quantum networks and is subject to continuous development and research. In this work, we present an example of classification for a set of higher dimensionality than the currently used solutions based on the MNIST or FASHION databases. Additionally, we show that working on raw data not transformed by, for example, PCA reduction or other advanced classical pre-processing techniques, very high classification quality can be achieved also without using any hybrid techniques. In the discussed solution, classification is performed by checking whether the obtained final state, or more precisely the probability distribution of the basis states superposition, is consistent with the appropriate state representing the given label. This comparison can be carried out using basic techniques such as Kullback-Leibler divergence or SWAP-Test, especially if we want the classification process to be realized solely in the quantum computation model without using any post-processing with classical techniques.