Convolutional neural networks for image object recognition and classification with large-scale and complex data

Abstract

The purpose of this study is to investigate the application of artificial neural networks for image recognition and classification tasks. Additionally, the research aims to explore the usage of convolutional neural networks in handling large-scale and complex data. These networks are designed to reduce memory storage requirements and hierarchically extract and aggregate features from input data, which is essential for efficient data processing. The research employs the use of convolutional neural networks and large-scale and complex data processing techniques, which are at the forefront of advancements in machine learning. The neural network is constructed and trained with large-scale and complex data, ensuring that the system is robust and capable of handling real-world datasets. The performance is evaluated using various configurations of convolutional and pooling layers, which are integral components of CNNs that help in feature detection and reduction of computational complexity. The study also involves the development of a library of methods based on the “.NET Standard 2.0” platform, which is a widely-recognized framework for building high-performance applications. Additionally, a window application using “.NET 6.0” and “WPF” platforms is developed, demonstrating the practical implementation of the research. The study explores the quality of the proposed convolutional neural network based on different configurations of convolutional and pooling layers and the size of the convolution filter. The best results are achieved when using 3 blocks of convolutional and pooling layers with a filter size of 3 x 3 pixels. The network achieves optimal accuracy in image object classification after being trained for 14 epochs. The findings demonstrate the effectiveness of the proposed convolutional neural network architecture and its ability to handle large-scale and complex data efficiently for image recognition and classification tasks. The study’s outcomes contribute significantly to the field of computer vision and offer a promising direction for future research in neural network optimization for image processing.