Waste management is essential as it ensures cleaner and healthier living conditions for both humanity and the environment. Type-based segregation, one of the pivotal roles of waste management, enhances recycling efficiency, reduces the usage of landfills, and minimizes contamination in recycling streams. This work is focused on using deep learning techniques, particularly CNNs, to divide garbage into six categories: trash, cardboard, paper, plastic, glass, and metal. This was implemented by fine-tuning pre-trained CNN models like VGG16 and MobileNet using a diverse and richly labeled garbage classification dataset. The preprocessing was conducted vastly for resizing, normalization, and data augmentation to further ensure robust training and reliable results. Accuracy, precision, recall, and F1-score are evaluation measures used to assess the model’s performance. VGG16 reached an impressive accuracy of 97%, surpassing MobileNet and an Ensemble model, which reached accuracies of about 96%. This approach not only sustains its progress in improving recycling processes but also contributes to reducing environmental degradation and overflow in landfills.