Volume : 9, Issue : 6, JUN 2023

NATIONAL CONFERENCE ON INNOVATIONS IN COMPUTING TECHNOLOGIES (NCICT'23)

GRANULATED RCNN AND MULTI-CLASS DEEP SORT FOR BIODEGRADABLE AND NON-BIODEGRADABLE SOLID WASTE DETECTION

P.GANESH, DR. K. HARIDAS

Abstract

Solid waste management (SWM) is a major problem for many urban local bodies (ULBs) in India, where urbanization, industrialization, and economic growth have resulted in increased municipal solid waste (MSW) generation per person. Waste management reduces greenhouse gas emissions and improves the quality of air and water, and the condition of any area affected by the waste. This can be overcome by separating the solid waste into two separate medium. The first one will collect the biodegradable waste such as paper, vegetable and fruit waste which can be converted in to energy by using Biomethanation plant. The second one will collect non-Biodegradable waste such as plastic, can, bottles which can be used for recycling purposes. The Biodegradable and Non-Biodegradable wastes separation should be implemented in automatic manner by applying granulated RCNN (G-RCNN) and multi-class deep SORT (MCD-SORT) to predict the type of waste item in accurate manner. The waste material detection has two stages: waste object or material localization (region of interest RoI) and classification. In deep convolution neural network, the G-RCNN is an improved version of the well-known Fast RCNN and Faster RCNN which helps to extract RoIs by incorporating the unique concept of granulation. Granulation with spatio-temporal information enables more accurate extraction of RoIs (waste object regions) in unsupervised mode. Compared to Fast and Faster RCNNs, G-RCNN uses (i) granules (clusters) formed over the pooling feature map, instead of its all feature values, in defining RoIs, (ii) only the positive RoIs during training, instead of the whole RoI-map, (iii) videos directly as input, rather than static images, and (iv) only the objects in RoIs, instead of the entire feature map, for performing waste object classification. All these lead to the improvement in real-time detection speed and accuracy. MCD-SORT is an advanced form of the popular Deep SORT. In MCD-SORT, the searching for association of objects with trajectories is restricted only within the same categories. This increases the performance in multi-class tracking.

Article : Download PDF

Cite This Article

-

Article No : 42

Number of Downloads : 1102

References

1. Wen-Tien T, “An Analysis of Operational Efficiencies in the Waste-to-Energy Plants of Kaohsiung Municipality", Resources 2019, July 2019.

2. Megha S, Chaudhari, Bharti P, Vaishali R “IoT based waste collection management system for smart cities”, 3rd International Conference on Computing Methodologies and Communication, 2019.

3. Shamin N, Fathimal M, Raghavendran R, “Smart Garbage Segregation & Management System using Internet of Things & Machine Learning”, 1st International Conference on Innovations in Information and Communication Technology, June 2019.

4. Pardini K, Rodrigres J.P.C, Kozlov A, “IoT Based Solid Waste Management Solutions:  A Survey”, Journal of Sensor and Actuator Networks, January 2019.

5. Zheng P, Xu S, Xindong, “Object Detection with Deep Learning: A Review”, IEEE Transactions on Neural Networks and Learning Systems, January 2019.