Generative Jamdani: A trailblazing research by AUST students

Staff Correspondent: Jamdani, the strikingly patterned textile heritage of Bangladesh is globally popular now but many of its motifs have been lost with time with demise of artisans as their lineage could not preserve all of it in just memory.
But think about a dataset of Jamdani motifs with which newer Jamdani patterns can be generated or some old ones with just a few hand strokes imitating the pattern itching your mind.
A research team of students from the Department of Computer Science and Engineering at Ahsanullah University of Science and Technology (AUST) recently developed such a prototype called Generative Jamdani using Machine Learning to bring that idea to reality.
The research aims at preserving the motifs of Jamdani and to get the country a step ahead in making the geometric motifs woven craftsmanship influential on modern textile and fine art industry and generating entirely new Jamdani motifs from a user’s input.
The team consists of four members, Humaira Ferdous Shifa, Md Tanvir Rouf Shawon, Raihan Tanvir and Susmoy Kar, who conducted this research as their undergraduate thesis under the supervision of Mohammad Imrul Jubair, Assistant Professor, of the CSE department of AUSt.
“Users can select any of the presets of motifs from the system we developed or draw rough strokes as input, and a complete motif mimicking the geometric structure of the real ones will be generated instantly,” said Humaira Ferdous Shifa .
“The team took advantage of a particular branch of Machine Learning, which is the Generative Adversarial Network. Their key idea was to make their system learn the patterns of authentic Jamdani designs by providing many original motifs”, said professor Imrul Jubair .
“They created a unique dataset named Jamdani Naksha, the first and only available dataset of Jamdani motifs in a digital format in Bangladesh which can be used for further research. It contains a large number of preprocessed images of Jamdani motifs collected from various authentic sources,” he added.
The team published the first research paper on this in the International Conference on Computer and Information Technology (ICCIT) 2020 and has made it to the top 50 at Unibator, the most extensive venture incubation competition launched by the Ministry of ICT, Bangladesh Hi-Tech Park Authority (BHTPA), and the Institution of Engineers, Bangladesh (IEB).
“If we generate an app for the dataset we developed, it can be used as a tool to generate innovative Jamdani motifs for not only cloths rather for any other crafts. It will also allow international or national businesses to flourish by connecting local artisans. The invention will help the fine art of Jamdani weaving resuscitate the industry by preserving the motif from further extinction while carrying on the legacy of the surviving motifs, “ said Humaira Ferdous Shifa, one of the team members.
The artificial production of Jamdani motifs will unveil a new source of inspiration for artists and designers by playing the role of an intelligent tool of imagination, she added.