DESIGN AND IMPLEMENTATION OF A DEEP LEARNING-BASED IMAGE CLASSIFICATION SYSTEM USING CONVOLUTIONAL NEURAL NETWORKS

The paper is a design and a realization of a deep learning-driven image classification system with the use of Convolutional Neural Networks (CNNs). The goal of the paper was to create a smart model that can properly classify images into the predefined classes by means of the automated feature extraction and learning. The system was created using the Agile approach and functionality, which meant that it was created in an iterative process where the subsequent sprints could be used to continuously improve the system by preparing datasets, training the model, testing, and evaluating the total output. The dataset was a collection of four categories which includes Cat, Dog, Bird, and Fish; the publicly available sources were chosen to use Kaggle and ImageNet. The images were pre-trained by resizing, normalization and augmentation methods such as rotation, flipping, scaling and adjustment in brightness to enhance generalization and reduce overfitting. The system implementation results showed high performance in all the categories, with the highest scores being in the Cat class (Precision: 0.94, Recall: 0.95, F1-Score: 0.945, Accuracy: 0.96), then the dogs, birds and fish respectively. The trained model was directly converted into a web application written in Flask and allowed one to upload and classify images in real-time, which confirmed the practicality and usability of the system. In general, the designed CNN-based image classification system was precise, effective, and scalable, which fulfilled its goal of a reliable image recognition in various categories. The study showcases the opportunities of deep learning as a field of computer vision and how CNNs can be successfully implemented in practice in fields like healthcare, agriculture, and security. It will be followed by future work that considers the use of transfer learning based on more advanced architectures and optimized to support mobile and edge deployment in order to increase its performance and flexibility further.