In the rapidly evolving food manufacturing sector, detecting defects in products such as broken eggs on conveyor systems is crucial for ensuring product quality and safety before delivery. Traditional inspection methods often need more efficiency and accuracy, leading to increased waste and compromised safety standards. Recently the studies have been invested in this issue to explore the ability of optimizing the current methods, however, the space left in this sector for improving efficiency and accuracy is open. In this paper, we address these challenges by leveraging advanced deep learning models to enhance detection capabilities. We utilize Convolutional Neural Networks (CNNs) for precise image recognition, Deep Neural Networks (DNNs) for analyzing complex patterns, Long Short-Term Memory (LSTM) networks for monitoring temporal data, and You Only Look Once (YOLO) version 11.0 real-time object detection algorithm for precise image detection. A hybrid CNN-LSTM model also integrates these approaches, significantly improving detection accuracy and efficiency. Our experiments demonstrate that YOLO11n-cls can achieve a highly competitive result against state-of-the-art accuracy and speed, reaching an accuracy of 99.537% and an inference time of 0.4ms per image. Our study contributes to developing more sustainable food manufacturing systems by reducing waste and improving operational efficiency. Our experimental results are highly promising at this first step as the accuracy rate of the best model reaches 94%.