Faster Region-Based Multi-Layer Convolutional Neural Networks for Cracked Detection in Eggshell Images

Automatic detection of cracked eggs is of great importance in terms of health. Today, the separation of cracked eggs is done by experts through observation. This process causes time loss and erroneous detections together with tiring. In this direction, a system based on Region-based Convolutional Neural Network (CNN) has been designed for the automatic detection of cracks in the egg surface. An original data set containing cracked eggs images were created for the training and testing phase of the proposed 16-layer CNN-based model. Cracked regions in 107 egg images using the MATLAB platform were labeled. Within the scope of experimental studies, an average precision of 95.69% was obtained by using the proposed model for cracked region detection. These results showed that the proposed computer-based system can be used to automatically separate cracked eggs in the food industry