<p>As laser engraving morphology defect detection becomes a key link in laser engraving technology, traditional detection methods have problems such as low efficiency and lack of accuracy. To meet the actual needs of laser engraving morphology defect detection, the study proposes a laser engraving morphology defect detection method based on transfer learning and DenseNet121. This study uses a laser engraving morphology defect image dataset, adopting transfer learning and histogram equalization methods to classify and enhance the images in the dataset. Moreover, a data partitioning strategy is used to divide the dataset into mutually exclusive subsets. Finally, DenseNet121 network structure is adopted for shallow and deep parameter fine-tuning. The outcomes revealed that the research method took 175ms and 286ms to detect cracks and porosity defect problems, respectively. In the real application test, the research method achieved 94.9% accuracy in the validation set at the 140th iteration. The real-time detection accuracy of the research method stabilized at 98.5% when the response time was 360ms. The above results show that the proposed laser engraving morphology defect detection method has better detection efficiency and accuracy. It enhances the detecting capability of laser engraving morphological defects and successfully addresses the issues of poor efficiency and inaccuracy of conventional techniques.</p>