Smart Garbage Classification

• Published in: Innovations in Intelligent Computing and Communication pp. 113 - 124
• Main Authors: Jain, Aviral, Khetriwal, Vidipt, Daga, Hitesh, Tripathy, B. K.
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing
• Series: Communications in Computer and Information Science
• Subjects: Convolutional neural network; Deep learning; EfficientNet; Recycling; Transfer learning
• ISBN: 3031232321; 9783031232329
• ISSN: 1865-0929; 1865-0937
• DOI: 10.1007/978-3-031-23233-6_8

Garbage recycling is a key aspect of maintaining our environment in good condition. Poor waste management has the potential to have a significant negative effect on the environment, on public health, and on the economy of the nation. The garbage must be separated into groups with similar recycling processes to make the recycling process much more effective and faster. Recycling, needless to say, is a very important task for all the countries. Garbage categorization is the most basic stage in enabling cost-effective recycling among the tasks required for recycling. Some of the well-known deep learning models for trash categorization include Densenet121, DenseNet169, InceptionResnetV2, MobileNet and Xception architecture. In this article, we propose a procedure to recognize single trash objects in images and categorize them into one of the recycling categories. A Convolutional Neural Network (CNN) with transfer learning is used. An analysis of the results obtained from the study shows that the EfficientNet-b0 CNN performs well under this scenario. The trash categorization problem for the target database can be efficiently handled using deep learning approaches as they offer a reliable foundation for image recognition with high consistency. Once a waste is put into the bin, the top compartment scans and predicts which type of waste it is, and then the respective lid of the bottom compartment opens, pushing the waste down. This prototype model of the system is proposed, which can be used in real-time implementation.