In the food and pharmaceutical industries, maintaining the internal quality of stainless steel pipelines is crucial to ensure product quality. The shaping of the pipeline’s during cutting and turning processes must meet stringent technical requirements. It is essential that the surface is free from defects such as slag, visible weld imperfections, overheating, lack of smoothness, and scratches. Improper execution of cutting and turning processes can result in defects like turning marks, which can lead to accumulation and adherence of medium transported through the pipelines. Over time, these accumulated residues within the pipeline walls can cause corrosion and compromise product quality in the food and pharmaceutical sectors, potentially resulting in products that fail to meet industry standards. This paper aims to develop a system using a USB camera combined with deep learning algorithm to detect and identify scratches inside the microbiological pipelines. The model is trained and tested on a dataset comprising 100 original images, augmented to 600 images using augmentation techniques. This dataset is split into 80% (480 images) for training and 20% (120 images) for evaluation. Two algorithms, YOLOv8 and SSD MobileNetV2, are employed to train and construct the model. Experimental results show that the YOLOv8 model achieves Precision, Recall, and F1-Score of 53.3%, 54.7%, and 54.0%, respectively. We then build an application to integrate YOLOv8 model for detecting scratches in microbiological piping.