MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641109853 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology on September 13, 2026, for A Deep Learning-Based Intelligent Waste Management System For Automated Waste Detection, Classification, And Segregation.
Inventor includes Dr. Sagar Yeruva.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: ABSTRACT [0021] Inefficient waste management remains a significant global issue, contributing to environmental degradation, resource wastage, and pollution. Traditional methods of waste sorting are often labour-intensive, error-prone, and unable to handle the complexity of diverse waste materials. The inability to accurately classify and separate waste hinders recycling efforts and sustainable disposal practices. This study introduces an automated solution based on the YOLOv5s model, trained on the TACO dataset, to detect and classify waste types with high accuracy. The model leverages optimizations such as the SGD optimizer with momentum and CIoU loss, providing a promising solution for real-time waste detection in urban and industrial environments. By automating the waste classification process, the system aims to improve recycling efficiency and contribute to more sustainable waste management practices. Its application could significantly reduce the reliance on manual sorting, streamlining waste management operations. [0022] While the results are promising, further refinements are necessary, particularly in improving the model's precision and recall. The existing model performed well but exhibited some challenges in handling specific waste types, highlighting the need for expanded and more diverse training data. Additionally, utilizing advanced data augmentation techniques could improve the model's ability to generalize across different types of waste. Although YOLOv5s strikes a balance between speed and accuracy, exploring more recent versions like YOLOv4 or YOLOv6 could provide even better performance. These newer models might offer improvements in both detection accuracy and computational efficiency, making them better suited for large-scale, real- time waste classification in diverse environments. Optimizing for better generalization could make the system more adaptable across various waste streams and real- world conditions.
Disclaimer: Curated by HT Syndication.
