Custom CNN for Sustainable Waste Segregation: Advanced Classification of Organic and Recyclable Material

This study presents a robust waste classification model leveraging a Custom Convolutional Neural Network (CNN) architecture to accurately categorize waste into Recyclable and Organic categories. The purpose of the study is to enhance waste management efficiency and environmental sustainability by automating waste sorting. Utilizing a dataset, containing 25,077 labeled images, the data was preprocessed to standardize image size and scale pixel intensities, facilitating optimal CNN performance. The data was split into training (90%) and testing (10%) sets, maintaining balanced representation across categories. The custom CNN architecture employs a multi-layered approach with dropout regularization to reduce overfitting, achieving a 90% classification accuracy. Key findings include high recall for Organic waste (99%) and precision for Recyclable waste (97%), underscoring the model's balanced performance across both categories. Evaluation metrics, including F1-scores and confusion matrix analysis, reveal robust categorization capabilities, although minor misclassifications in Recyclable items suggest potential areas for improvement. This research significantly advances waste classification by demonstrating an effective model that could be integrated into IoT-enabled waste bins for automated sorting. Such a system would support sustainable waste management practices, contribute to smart city initiatives, and minimize human intervention in waste sorting, promoting eco-friendly solutions for waste management.