Exploratory Analysis and Dataset Creation of Qajar Carpet Designs and Motifs: A Data-Driven Approach with a Focus on Machine Learning

This research primarily seeks to conduct an exploratory, data-centric examination of carpet motifs and designs from the Qajar era, employing modern machine learning techniques to achieve this goal. By collecting images of Qajar carpets from renowned museums such as the Victoria and Albert Museum, the Metropolitan Museum of Art, and the Hermitage Museum, a dataset comprising ۸۰۰ metadata entries (۲۰۰ original images and ۶۰۰ augmented versions generated using grayscale, Laplacian, and Gabor filters) was created. The study examines the statistical distribution of carpet designs, visual features, design patterns, dimensions, and weaving regions. The “Seven-Face” classification—Mohramat, Ghabi, Vagirehii, Lachak-Toranj, Mehrabi, and Afshan—serves as the main analytical framework, revealing that Afshan (۳۰%) and Lachak-Toranj (۲۳%) designs are predominant. In comparison, ۴۱% of the carpets incorporate pictorial elements influenced by Western cultural aesthetics. Kerman and Kashan emerge as primary production centers; however, challenges such as missing regional information (۲۸.۵%) and incomplete collection data (۴۳.۵%) persist. This research demonstrates that integrating computational methods with art-historical studies can deepen our understanding of the evolution of design, stylistic origins, and regional distinctions in Qajar carpets, thereby paving new pathways for interdisciplinary research in the field of Iranian traditional arts.

کلیدواژه ها:

Carpet ، Design and Motif ، Labeling ، Machine learning ، Qajar period