Eye Recognition by YOLO for Inner Canthus Temperature Detection in the Elderly Using a Transfer Learning Approach

Abstract

Early detection of physical frailty and infectious diseases in seniors is important to avoid any fatal drawback and promptly provide them with the necessary healthcare. One of the major symptoms of viral infections is elevated body temperature. In this work, preparation and implementation of multi-age thermal faces dataset is done to train different “You Only Look Once” (YOLO) object detection models (YOLOv5,6 and 7) for eye detection. Eye detection allows scanning for the most accurate temperature in the face, which is the inner canthus temperature. An approach using an elderly thermal dataset is performed in order to produce an eye detection model specifically for elderly people. An application of transfer learning is applied from a multi-age YOLOv7 model to an elderly YOLOv7 model. The comparison of speed, accuracy, and size between the trained models shows that the YOLOv7 model performed the best (Mean average precision at Intersection over Union of 0.5 (mAP@.5) = 0.996 and Frames per Seconds (FPS) = 150). The bounding box of eyes is scanned for the highest temperature, resulting in a normalized error distance of 0.03. This work presents a fast and reliable temperature detection model generated using non-contact infrared camera and a deep learning approach.

PhoGAD: Graph-based Anomaly Behavior Detection with Persistent Homology Optimization

“…In the scenarios mentioned in Section 3.1, spam detection stands out as a specific case. Certain studies have employed complex models to extract textual features from emails [11,38]. However, these methods frequently encounter challenges in capturing the interrelations between emails and demand domain-specific knowledge.…”

“…To ensure effective filtering of adjacent edges, we introduce the convolution parameter penalty based on node similarity, which is denoted by 𝛽 in Equations ( 10) and (11). This allows the representations of edges connecting similar types of nodes to fuse more quickly, thereby facilitating the identification of localized anomalies.…”

Investigating Evasive Techniques in SMS Spam Filtering: A Comparative Analysis of Machine Learning Models

“…Almeida et al compared several ML classifiers and found SVM performed well, however, they only used word frequency as a feature and did not evaluate DL models [4], [41]. Similarly, Gupta et al compared 8 ML clas- 747 [32], [33], [7], [34], [35] NUS SMS Corpus [31] 2015 67,063 Nil [36], [37], [38] SpamHunter [14] 2022 25,889 947 [39], [40] sifiers, including SVM and CNN, using only TF-IDF features and only evaluated conventional ML and one DL classifier [6]. Roy et al [42] compared CNN and LSTM with traditional two-class classifiers for SMS spam detection using multiple model stack structures and found that CNN and LSTM performed better.…”