The MNIST and Fashion MNIST datasets are used in this work to create a Convolutional Neural Network (CNN) model specifically designed for picture classification applications. CNNs have attracted a lot of interest because they are good at identifying complex patterns in visual data. Our research attempts to classify handwritten numbers and fashion items with great accuracy by utilizing these capabilities. With around 98% accuracy on the MNIST dataset and about 90% accuracy on the Fashion MNIST dataset, the model exhibits impressive performance after being painstakingly trained and verified. Using Python's PyTorch framework, the architecture is developed, utilizing deep learning tasks' flexibility and efficiency. We improve the network parameters through rigorous testing to maintain computational efficiency while maximizing classification accuracy. Our results highlight the CNN's promise for practical uses in object recognition and classification as well as its resilience when processing a variety of picture datasets. Furthermore, our findings demonstrate how effective CNN is in classifying images compared to more conventional approaches, presenting encouraging directions for future computer vision research and advancement. By highlighting the CNN's capacity to meet the expanding needs of AI applications, this work opens the door for developments in automated image analysis, object detection, and pattern recognition, among other subjects. CNNs become increasingly important as computing systems advance because they can handle the difficulties presented by large, varied datasets and a variety of application domains.