Abstract: The digital management of radiographic films for non-destructive testing of oil and gas pipeline welds is a crucial component of pipeline integrity management. To address the issues of heavy manual workload and low efficiency in identifying mark information on weld radiographic films, as well as the challenges posed by misaligned, overlapping, obscured, or inverted signature marks that make traditional methods difficult to apply, an intelligent recognition method for pipeline weld radiographic film marks—YOLO-MPW is proposed. Based on the YOLOv11 baseline model, the design of the Convolutional Gated Linear Unit Transformer (CF-CGLU) and the lightweight detection head (LDH), along with the introduction of the CARAFE operator, enhances the model's responsiveness to important features and improves the utilization of semantic information in feature maps, while significantly reducing the model's parameters and complexity. Using a dataset of radiographic images from welds in an oil and gas pipeline in southern China, the algorithm was trained and validated. The results show that the method achieves an mAP@0.5 of 98.6%, representing a 2.5% improvement over the baseline model, while reducing the number of parameters by 17.2% and computational load by 18.2%. The YOLO-MPW model balances accuracy and lightweight design, effectively achieving fast and accurate recognition of mark characteristics on pipeline weld radiographic films.