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

Abstract

In daily life, snail classification is an important mean to ensure food safety and prevent the occurrence of situations that toxic snails are mistakenly consumed. However, the current methods for snail classification are mostly based on manual labor, which is inefficient. Therefore, a snail detection and classification method based on improved YOLOv7 was proposed in this paper. First, in order to reduce the FLOPs of the model, the backbone of the original model was improved. Specifically, the original 3×3 regular convolution was replaced with 3×3 partial convolution, and the Conv2D_BN_SiLU module in the partial convolution was replaced with the Conv2D_BN_FReLU module. FReLU could enhance the model's representational capacity without increasing the number of parameters. Then, based on the specific features of snail images, in order to solve the problems of small and dense targets of diverse shapes, a receptive field enhancement module was added to the head to learn the different receptive fields of the feature maps and enhance the feature pyramid representation. In addition, the CIoU was replaced with the WIoU to make the model pay more attention to targets at the edge or difficult-to-regress accurate bounding boxes. Finally, the images of nine common types of snails were collected, including the Pomacea canaliculata, the Viviparidae, the Nassariidae, and so on. These images were then labeled using LabelImg software to create a snail image dataset. Experiments were conducted based on the dataset, and the results showed that the proposed method demonstrated the best performance compared to other state-of-the-art methods.

Keywords

• food safety
• snail species identification
• yolov7
• frelu
• pconv
• rfem
• wiou