Deep Learning Approaches for Egg Crack Detection: A Comparative Study Using CNN, ResNet, and Xception Models

In the food industry, egg crack prediction is crucial for ensuring the quality as well as the safety of eggs before they are eaten. It allows for real-time egg sorting and monitoring, improving the production process’s productivity and reduces the cost-effectiveness. Images of both damaged and undamaged eggs were collected, and the dataset was significantly expanded through data augmentation techniques. By expanding the dataset, this method improved the model’s capacity to forecast egg breaks with accuracy. Preprocessing was performed using data augmentation, scaling, and normalization on the gathered photos of both injured and undamaged eggs. These processed images were then fed into various deep learning models, including CNN, ResNet, and Xception. ResNet achieved the highest accuracy of 98.75 percent in predicting egg cracks. These results underline the value of dataset augmentation in enhancing model accuracy and demonstrate the efficiency of deep learning models in quality control within the egg business. Furthermore, the study also shows how crucial dataset augmentation is to raising machine learning models’ accuracy in identifying egg breaks. Future research on the creation of efficient machine learning models for early identification as well as quality control for egg cracks can build on the results of this work.