Deep learning (DL) is an influential technique for analyzing large-scale images. It includes the use of neural network (NN) with several layers to extract complex features from images and classify them into different types. DL models are accomplished of handling large quantities of data and can effectively identify patterns and relationships in images. This makes them well- suited for jobs such as image recognition, object discovery, and image classification in various fields, together with healthcare, finance, and transport. Therefore, DL plays a crucial part in analyzing and taking large- scale images. In recent years, there has been a noticeable lack of alternative methods based on (parametric) Bayesian implication. The intricacies and specificity of the data in remote sensing situations might make it difficult to assume a precise previous distribution of the data. In this regard, the developing discipline of nonparametric Bayesian methods provides an appropriate theoretical frame work to address the problem of classifying remote sensing images. In order to categorize handwritten digits and color photographs, respectively, the study e entailed training and trying Convolutional Neural Networks (CNNs) on two datasets: MNIST and Cifar-10. The purpose of the study was to find out how well DL models work for tasks involving image classification. The study found that by exhausting GPUs for computation, CNNs can extremely cut down on computation time and are very successful at extracting complicated information from input photos. The results of this study show that DL models have the ability to achieve high accuracy rates in picture classification jobs, which has important inferences for the field of image classification.