Visual Convergence of Industrial Heritage Cultural Genes and Potential Cross-Cultural Readability: A Dual-Task Learning Framework for Cross-Platform User-Generated Content

Industrial heritage contains industrial value, region-specific remains, and cultural identity; however, the extent to which its visual representations support a multicultural reading remains an empirically unresolved question. This study explores the potential of visual encounters and multicultural readings in the “cultural reproduction” of industrial heritage sites through a dual-task learning process using cross-platform user-generated content (UGC). Using UGC images of the Wuhan Yangtze River Bridge in China and the Forth Bridge in the UK, the study examined visible cultural genes in six functional categories and spatial visual attributes in six categories. The proposed system achieved a Macro-F1 score of 0.807 for cultural genes, a mean absolute error (MAE) of 0.117 for spatial visual attributes, and an evidence consistency rate (ECR) of 0.836. Compared with the shared Vision Transformer dual-task model with no evidentiary limitations (M2), the full model (M3) improved Macro-F1 by 0.028, reduced MAE by 0.015, and increased ECR by 0.074. This comparison provides an internal reference for interpreting model performance rather than an external benchmark against general performance standards. The analysis shows shared visual patterns along with case-specific differences in the representation of industrial heritage cultural genes. These findings identify candidate visual cues for future research on cross-cultural interpretation. However, this study examines computer-presented images rather than directly measuring whether spectators from different cultures understand the cultural significance of the two bridges. Consequently, the findings provide evidence for the congruence of a particular visual narrative and the foundation for further testing the potential for cross-cultural reading, rather than direct evidence for cross-cultural understanding.