CSU-Carig
Rafael, Reanne Joy G., author.
Detection of molds in BPI-Pn9 peanut seeds using image analysis and machine learning [manuscript] / Reanne Joy G. Rafael, Melanie S. Capalungan, B-Jay L. Daguio, Isaac A. Balbuena. - ©2022. - xvii, 73 pages ; 28 cm.
Thesis (B.S.) -- Cagayan State University, 2022.
Includes bibliographical references.
Peanuts with molds may contain aflatoxin, a highly carcinogenic substance that represents a health hazard. Most routine tests for molds in food samples like peanuts use conventional colony count methods rather than specific detection techniques, which require 5-7 days to achieve a result. In contrast, rapid detection methods can reduce this by 50% or more. This study aimed to identify mold-contaminated peanuts using image analysis and machine learning. Samples of peanut seeds were sterilized and subjected to inoculation, and a laboratory test was performed in accordance with FDA and DOST to confirm the presence of molds at 1.3×102 CFU/g. Detailed images of with and without molds peanuts were captured. The detection of molds on peanut surfaces was based on the "you only look once (YOLO) v5" principle. In this context, a dataset of 828 preprocessed and augmented images was used to detect mold-contaminated peanuts. The dataset was trained using the YOLOV5 algorithm. The precision and recall performance of the YOLOV5 model were determined. The YOLOV5 model performed well in 150 epochs, in a resolution of 640x640 pixels, with a confidence threshold of 0.75 and an IoU threshold of 0.50. It achieved a precision rate of approximately 99% with a recall rate of 100%.
Keywords: Image Analysis, Machine Learning, Molds, Peanuts, yolov5.
Agri 0156 / 2022 c.1
Detection of molds in BPI-Pn9 peanut seeds using image analysis and machine learning [manuscript] / Reanne Joy G. Rafael, Melanie S. Capalungan, B-Jay L. Daguio, Isaac A. Balbuena. - ©2022. - xvii, 73 pages ; 28 cm.
Thesis (B.S.) -- Cagayan State University, 2022.
Includes bibliographical references.
Peanuts with molds may contain aflatoxin, a highly carcinogenic substance that represents a health hazard. Most routine tests for molds in food samples like peanuts use conventional colony count methods rather than specific detection techniques, which require 5-7 days to achieve a result. In contrast, rapid detection methods can reduce this by 50% or more. This study aimed to identify mold-contaminated peanuts using image analysis and machine learning. Samples of peanut seeds were sterilized and subjected to inoculation, and a laboratory test was performed in accordance with FDA and DOST to confirm the presence of molds at 1.3×102 CFU/g. Detailed images of with and without molds peanuts were captured. The detection of molds on peanut surfaces was based on the "you only look once (YOLO) v5" principle. In this context, a dataset of 828 preprocessed and augmented images was used to detect mold-contaminated peanuts. The dataset was trained using the YOLOV5 algorithm. The precision and recall performance of the YOLOV5 model were determined. The YOLOV5 model performed well in 150 epochs, in a resolution of 640x640 pixels, with a confidence threshold of 0.75 and an IoU threshold of 0.50. It achieved a precision rate of approximately 99% with a recall rate of 100%.
Keywords: Image Analysis, Machine Learning, Molds, Peanuts, yolov5.
Agri 0156 / 2022 c.1