<p>The increasing adoption of selective laser melting (SLM) in additive manufacturing has unlocked new capabilities in producing high-precision metal components with complex geometries. However, the stability and quality of individual print tracks remain a major challenge due to thermal-induced deformations and variability in laser parameters. This study proposes a deep learning-based approach using convolutional neural networks (CNN) to classify 3D printed line segments captured through optical microscopy into high- and low-quality categories. The original dataset, comprising only 422 holographic images, was systematically partitioned and expanded through data segmentation techniques to create larger labeled datasets suitable for model training. The proposed CNN architecture achieved high classification accuracy on two augmented datasets, reaching 99.01% on the larger 5064-image set. Comparative evaluations with traditional machine learning models and alternative CNN architectures further validate the superiority of the proposed approach. This research contributes a robust, scalable solution for automated print line quality assessment in SLM, with the potential to reduce trial-and-error costs and enhance real-time print parameter optimization in industrial environments.</p> Graphical abstract <p></p>