A snail species identification method based on deep learning in food safety

Area of Science:

• Computer Vision
• Artificial Intelligence
• Food Safety Technology

Background:

• Manual snail classification poses inefficiencies and risks in ensuring food safety.
• Accurate identification of snail species is crucial to prevent consumption of toxic varieties.

Purpose of the Study:

• To develop an efficient and accurate automated snail detection and classification system.
• To improve upon existing deep learning models for enhanced snail identification.

Main Methods:

• An improved YOLOv7 model incorporating partial convolution and FReLU activation for backbone optimization.
• Integration of a receptive field enhancement module and WIoU loss function to address challenges with small, dense, and varied snail targets.
• Creation of a comprehensive snail image dataset comprising nine common species, including Pomacea canaliculata, Viviparidae, and Nassariidae.

Main Results:

• The enhanced YOLOv7 model demonstrated superior performance compared to other state-of-the-art methods in snail classification tasks.
• Model improvements successfully reduced computational load (FLOPs) while enhancing representational capacity.
• The WIoU loss function improved the model's ability to accurately detect challenging targets, such as small or edge-case snails.

Conclusions:

• The proposed improved YOLOv7 method provides an effective solution for automated snail classification.
• This technology has significant potential to enhance food safety protocols and reduce risks associated with misidentifying edible and toxic snails.
• The study highlights the efficacy of targeted deep learning model modifications for specialized image recognition tasks.