Class imbalance is a major challenge in machine learning, particularly when working with datasets where one class significantly outweighs others. We introduce a smarter way to handle imbalanced datasets by combining fuzzy logic with Cost-Sensitive Learning (CSL) in the ResNet-50 architecture. This approach dynamically adjusts class weights during training, helping the model recognize minority classes more accurately while maintaining overall performance. Techniques such as resampling and CSL, including fuzzy logic, have been extensively studied. The incorporation of fuzzy logic in Cost Sensitive Learning allowed the system to handle uncertainty in data more effectively, further enhancing classification performance. However, most methods focus on either altering the data or modifying algorithms separately. Using the ResNet-50 architecture integrated with CSL, our approach dynamically adjusts misclassification costs during training, enabling balanced and robust feature learning across all classes while preserving the original data distribution. The proposed methodology minimizes the need for extensive preprocessing, making it highly practical for real-world applications. Applied to image classification tasks such as garbage categorization, where class imbalance is a common problem, the proposed model achieved an accuracy of 97.06%, significantly exceeding the prior benchmark of 94.63%. Experiments on 10 imbalanced datasets demonstrate the effectiveness of framework, scalability, and potential for real-world image classification applications. This balance between accuracy and computational efficiency makes the framework suitable for deployment in resource-constrained environments.