Publication: E. Coli colony detection and counting using deep learning approach
Loading...
Date
2023-07-01
Authors
Muhamad Syahmi Bin Johar
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Escherichia coli (E. coli) has been widely acknowledged as a pathogenic threat which causes detrimental effect on human health. Early detection of its presence especially in food and water is deemed vital as it can be a mitigation strategy to prevent its widespread. Detection of E. coli via sample-ready petrifilm plate is one of the relevant options for the public to handle this issue. However, the shortcoming in quantification process of this method, which now relies on manual counting is still yet to be tackled. By utilizing YOLOv4 deep learning algorithms which have the capability to detect and quantify the E. coli present in the petrifilm plate, this method has been further improved and more convenient for public usage. The developed E. coli detection model was tested before and after being equipped within a Raspberry Pi. The test before the model is equipped within a Raspberry Pi was with the intend of evaluating the model overall detection performance while the test after the model is equipped within a Raspberry Pi was done to assess the performance of the device’s system when running the detection. The results showed that the developed E. coli detection model managed to produce a high detection performance with precision of above 92%, recall above 84%, and F-1 score above 88%. The device’s system performance was assessed mainly through the inference speed when running a detection where the results showed that an inference time average at under 4 seconds was managed to be achieved using YOLOv4-tiny. This project can be seen as the first step in making the detection for presence of E. coli in the specialized petrifilm more portable and accessible to the public.