DEVELOPMENT AND TRAINING OF A NEURAL NETWORK AUTOENCODER MODEL FOR VISUAL DATA COLORIZATION

Abstract

This paper focuses on the development and training of a convolutional autoencoder model for the task of automatic colorization of grayscale images. The relevance of the study is driven by the rapid integration of artificial intelligence techniques into the field of visual data processing and the growing need for improved accuracy in automated color information reconstruction. The objective of the research is to design a neural network architecture capable of performing high-precision color restoration while preserving the spatial structure of the input image. The study addresses key tasks such as theoretical analysis of computer vision methods, classification of neural network architectures, and practical implementation of the model using the Keras library. The training conducted on a dataset of landscape images resulted in an accuracy of 82.5%, with visually satisfactory colorization quality. The findings confirm the effectiveness of the proposed architecture and demonstrate the potential of convolutional autoencoders for color information recovery. The developed model can be applied in digital restoration projects, content generation, and educational settings.