To address the resource consumption issue of obtaining precise paired samples in existing fully supervised learning, while also considering the quality of network map generation, a novel semi-supervised online map generation model based on generative adversarial networks was proposed, which aimed to realize the direct generation of intelligent remote sensing images into network maps by using only a few precisely matched data and a large amount of unpaired data. In addition, a semi-supervised learning strategy based on transformation consistency regularization and sample enhanced consistency was designed, which overcomed the inconsistency problem caused by imprecise paired data and derives better generalization performance of the model. Adequate comparison experiments were conducted on different map datasets. The generated online maps outperform the competing methods on the quantitative metrics and visual quality, which validate the effectiveness and speed of semi-supervised network map generation methods.