Solid waste management has emerged as one of the most significant issues confronting the globe today. It involves waste generation, collection, transportation, treatment, and disposal. Accurate waste classification is essential for effective collection, disposal, and recycling. Numerous deep learning and machine learning algorithms are available for classifying images, with the most popular being convolutional neural networks. Pretrained models, a subset of deep learning approaches, have emerged as powerful tools for tackling image classification tasks. While researchers have utilized pretrained models like VGGNet and ResNet for waste classification tasks, these models often face challenges such as limited feature extraction capabilities for diverse visual patterns, higher computational costs, or difficulty in training deeper architectures. This study employs the pretrained InceptionResNetV2 model, enhanced through transfer learning, to address these issues through its hybrid design, combining the multi-scale feature extraction of Inception modules with the training stability of residual connections. The model achieved an outstanding accuracy of 99.89% on the TrashNet dataset, comprising six waste categories (glass, metal, cardboard, paper, plastic, and trash) and 2527 images. The results indicate the effectiveness of transfer learning in optimizing complex deep learning models for small yet diverse datasets. The ability of the InceptionResNetV2 model to generalize across various categories highlights its robustness, making it a viable solution for real-world waste management applications. This study emphasizes the power and efficiency of utilizing pretrained models for waste classification, paving the way for more advanced and automated solid waste management systems.