| Autor: |
Agarwal, Shivaank, Gudi, Ravindra, Saxena, Paresh |
| Zdroj: |
Transactions of Indian National Academy of Engineering; September 2022, Vol. 7 Issue: 3 p739-751, 13p |
| Abstrakt: |
This paper proposes a computer vision based image classification approach to classify plastic wastes based on their resin identification code, to enable an efficient recycling of these wastes. While classification approaches to deal with waste plastic can be developed for known kinds of plastic, there is a need to also accommodate the challenges associated with the diversity of plastic wastes. In this paper, we propose machine learning methods for the following two problems. First, to classify a plastic waste into one of the known categories of the waste and second, to identify plastic wastes that do not belong to any known categories of the waste. We propose the use of one-shot learning techniques based on siamese and triplet loss networks to extract features for the plastic waste images. We then use the supervised and unsupervised dimensionality reduction techniques on the extracted features to classify the images into the known or unknown class of plastic wastes. If the plastic waste does not belong to an unknown category, we identify the resin code category it belongs to. We demonstrate a high accuracy of the classification approach to perform the task of plastic waste segregation by validation on the WaDaBa database that contain images of such plastic wastes (Bobulski and Piatkowski in Pet waste classification method and plastic waste database-wadaba. In: International conference on image processing and communications, p 57–67, 2017). |
| Databáze: |
Supplemental Index |
| Externí odkaz: |
|
Full text from SpringerLink