This study proposes a non-contact automatic measurement system for traditional fiber artifacts, specifically focusing on plain woven textiles. Due to the organic nature of fiber artifacts, which makes them susceptible to rapid degradation through oxidation and microbial activity, prompt and precise documentation is essential. The system employs high resolution images captured using a portable optical microscope that are subsequently enhanced by applying R-CUGAN and ESRGAN (Wang, X., Yu, K., Dong, C., Loy, C.C.: ESRGAN: Enhanced super-resolution generative adversarial networks. In: Proceedings of the European Conference on Computer Vision (ECCV) Workshops, pp. 0–0 (2018)) super resolution techniques. The images are meticu-lously labeled at the pixel level using the Roboflow tool to distinguish warp, weft, scale, and exclusion regions, and then analyzed using a deep learning U-Net model. The enhanced U-Net, incorporating an encoder–decoder architecture with skip connections and residual learning, effectively segments the fine patterns within the artifacts. The automatically derived data on fiber thickness and density exhibit a high correlation with manual measurements. These results suggest that the proposed method can provide critical foundational data for artifact preservation, restoration, and the creation of digital archives in the field of digital heritage.