Detection And Counting Of E. Coli On Specialized Test Piece Using Yolo v4
Loading...
Date
2022-07-15
Authors
Teoh, Mynn Wei
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Supplying clean, safe and drinkable water is still one of the on-going issues faced by
the world. To date, people around the still contract sickness and diseases related to
unsanitary water. One of the most common sicknesses is diarrhoea and the main
contributor to it is Escherichia Coli or in short, E. Coli. E. Coli is a bacterium
commonly found in environment and if consumed in moderate and high amounts,
may lead to critical illness and death. Therefore, there is a dire need for vision
automation in detection of E. Coli bacteria. To date, the process of identify the quality
of water is still not accurate and is time consuming. The quality of water is measured
by colony forming unit per 100mL or in short CFUs/100mL. Only counting of the
colonies is possible to obtain that desired value, which even until today, is still
counted by sight. This leads to inaccurate E. Coli colony reading and inappropriate
water treatment procedures. The study includes the usage of machine learning
capabilities to detect and count the colony present on the test piece. The sample
images obtain from the laboratory is captured under ideal lighting condition and later
augmentation process was carried out. The processed images are then annotated using
Label Studio and later trained using YOLO v4, an object classifier network that
employs Convolutional Neural Network (CNN). The network is being trained to pick
up presence of E. Coli on the mentioned test piece and provide user the quality of
water based on the CFU. The results showed that with only 50 test piece sample
images, the model achieve a mAP accuracy of approximately 91%, IOU score of 0.82
and an average loss of 0.2588. During the test phase, this work recorded a precision of
0.9279 ± 0.04195, recall of 0.9474 ± 0.01831 and F-score of 0.9351 ± 0.02718.
This research is the first step to automate the E. Coli detection and counting process
and create a change to world.