Detection of mold on the food surface using YOLOv5

Area of Science:

• Food Science and Technology
• Computational Biology
• Mycology

Background:

• Mold contamination on food surfaces poses significant risks to food safety and quality.
• Accurate and efficient detection methods are crucial for preventing the distribution of moldy food products.
• Traditional methods for mold detection can be time-consuming and labor-intensive.

Purpose of the Study:

• To develop and evaluate a novel approach for detecting mold on food surfaces.
• To utilize the You Only Look Once (YOLO) v5 algorithm for automated mold identification.
• To compare the performance of YOLO v5 against previous versions (YOLOv3, YOLOv4) for mold detection.

Main Methods:

• A dataset of 2050 images featuring mold on various food surfaces was curated from laboratory and internet sources.
• The pre-trained YOLOv5 algorithm was employed to train the model on the created dataset.
• Laboratory tests were conducted to confirm the presence and nature of the detected organisms as mold.

Main Results:

• The YOLOv5 model demonstrated superior performance compared to YOLOv3 and YOLOv4.
• Achieved high precision (98.10%), recall (100%), and average precision (AP) of 99.60% for mold detection.
• This marks the first application of the YOLOv5 algorithm specifically for detecting mold on food surfaces.

Conclusions:

• The proposed YOLOv5-based model is highly effective in recognizing diverse types of mold on food.
• This technology offers a promising solution for rapid and accurate food mold detection.
• Future research is underway to extend the model's capability to identify specific mold species.