Publication Type : Conference Paper
Publisher : Springer
Source : International Conference on Micro-Electronics and Telecommunication Engineering
Url : https://link.springer.com/chapter/10.1007/978-981-16-8721-1_61
Campus : Coimbatore
School : School of Computing
Year : 2022
Abstract : In this article, deep learning architecture has been used to detect anomalies on thermal images of perishable items like fruits and vegetables. Three experiments have been conducted, and their performances have been studied. In first experiment, two different deep learning models have been applied for classifying presence or absence of anomaly and Deep-CNN (Deep-Convolutional Neural Networks) performs well. In second experiment, YoLoV4 and YoLoV5 have been used to classify and localize the anomalies; YoLoV5 localizes the anomalies effectively with a mAP (Mean Average Precision) of 95.67%. In third experiment, YoLoV5 and Faster-RCNN (Faster Region Based Convolutional Neural Network) have been trained on the dataset in which each image contains multiple objects with and without anomalies to detect and localize them; Faster-RCNN gives mAP of 96.55% for detecting and localizing anomalies. From the experiments, it is observed that proposed deep learning models can be used for anomaly detection in thermal images.
Cite this Research Publication : Ramyapriyanandhini, G., T. Bagyammal, Latha Parameswaran, and Karthikeyan Vaiapury. "Anomaly detection in thermal images of perishable items using deep learning." In International Conference on Micro-Electronics and Telecommunication Engineering, pp. 647-659. Singapore: Springer Nature Singapore, 2021.