The current biological 3D printer relies on electron microscopes to observe printed results and uses the three-point circle method to calculate the area for evaluating the product uniformity. However, this approach has notable issues, including time consumption, subjectivity, and large discrepancies with the real situation. Thus, an improved YOLOv8-Seg model was proposed for contour recognition of the printed products. The Adam optimizer was employed, and the training parameters of the original YOLOv8-Seg model were adjusted to ensure that the model’s confidence level for product recognition was mostly above 0.94. The trained model achieved an mAP50 of 99.5% and an mAP50?90 of 98.4%. Subsequently, data images in which a 500 μm scale bar was pre-placed were used to directly calculate the recognized contour area. Additionally, an algorithm for evaluating the similarity of the recognized contour to a circle was incorporated to optimize the metric for assessing product uniformity. The optimized algorithm showed that the difference between the recognized contour and the real contour was less than 0.25%. Finally, variance coefficient (CV) and circularity analyses were performed on the contour areas obtained by different methods. Results indicated that when the CV was less than 20% and the circularity was greater than 0.65, the printed product could be considered uniform.