Deep learning models have proven to be remarkably effective in a variety of computer vision applications in recent years, especially picture categorization. Among these models, ResNet-50, a convolutional neural network architecture, has garnered significant attention for its ability to effectively handle deep network architectures and mitigate the vanishing gradient problem. In this work, we investigated ResNet-50’s potential for image classification using the CIFAR-10 dataset, a popular benchmark dataset for assessing image classification techniques. This study’s main goal is to improve ResNet-50’s performance on the CIFAR-10 dataset by using a variety of optimization strategies and architectural changes. Firstly, we present a comprehensive overview of the ResNet-50 architecture and its underlying principles. Subsequently, we delve into the CIFAR-10 dataset, highlighting its characteristics and challenges.