Eggshell crack detection using deep convolutional neural networks

The accurate detection of cracks on eggshells is quintessential to provide consumers with safe and quality eggs. This study proposes a novel approach to automatically detect these eggshell cracks using a convolutional neural network (CNN). The CNN model was trained on image patches extracted from the egg images with different batch sizes. Inferences upon testing showed that the model trained on a batch size of 64 gave the best results. The model also outperformed the support vector machine (SVM) classifiers trained on the histogram of oriented gradients (HOG) and local binary pattern (LBP) features with precision, recall, accuracy, false positive rate (FPR) and area under the curve (AUC) of 98.28%, 92.41%, 96.92%, 0.81% and 99.53%, respectively in classifying image patches. In addition, the proposed CNN model surpassed the SVM models in classifying images of eggs with an accuracy of 95.38%.