As international students from countries with distinct histories and cultures, my project partner and I were interested in raising awareness for cultural textiles. This interest led us to our next question — can we highlight the uniqueness of cultural textiles through a data-driven process? As a result, we have created an exploration tool that highlights similarities and differences of traditional patterned textiles around the world through machine learning.
Since our tool's main use case is to raise awareness and advocate, the main user is envisioned to be the general public. Additionally, this tool will be helpful for fashion designers finding inspiration or art historians to kickstart research.
Two types of data collection methods were used to find data for our data visualization tool: physical books and online collections. Below are the resources we consulted.
In order to group our data by similarity, we used a deep learning model called scikit-learn. The algorithm utilized k-means clustering to sort each of our 1500+ textile swatches into a group. Because we didn't want groups that were too big, we chose the number of groups to be 100, so that each group would have 15 textile swatches on average. With the results, we designed an interactive high fidelity mockup on Figma.
We determined four main features for our tool: a map view, an archive page where all of the data can be clustered by different metrics, a find feature, and a share feature.
Flow of tool
Three key design principles drove our design — visual, exploratory, and diverse. We wanted our product to communicate the diverse, soft, and curious nature of exploring global textile patterns.
As can be seen in the beginning stages of the design, we were going for a futuristic look for the visual design at first. However, as we re-evaluated the design principles and what emotion we wanted to evoke through our visualization, we decided that we wanted to go for a softer and more traditional look.
Initial sketch
Digital iteration of initial sketch
Mid-fi design after pivot
Final design
A tool of discovery and inspiration for all.
Check out our two-minute video to view micro-interactions and get a walkthrough of the tool.
The main map view of the tool showcases traditional textiles from each region in the form of a hexagon. The user can interact with the map as they would with any other digital map. When the user hovers over a hexagon, they can view details of that pattern and other relevant information.
The tool allows the user to view all of the textiles in the tool database based on different categories, one of them being pattern. These clusters are grouped by a deep learning model, which shows an interesting angle of how AI views fabric patterns.
When the user clicks on one of the clusters, they get a focused view of that cluster, which shows more details about each textile swatch. The user can explore how cultures close and far vary or converge on textile design.
This tool allows users to compare their own textile swatch with textiles in our tool's database, which can be a great kickstarter for further research. Users can also add their own data to our database, which can further enrich our tool and benefit the public.
Once the user uploads a mystery textile swatch to our tool, the tool utilizes a deep learning model to identify similar textiles from its database. The results in the screen above are actual results from running the swatch through the model, which shows the potential of harnessing the power of AI to textile research and identification.
"I think it is awfully cool. It does exactly what a fashion designer would want, and the find feature is a great research tool for art historians who are looking to further their investigations into certain pieces. The link to the owner’s page will be useful in helping them reach out to find out specific details about the piece like yarn density and more."
— Curator at Harvard Art Museums
This project allowed me to appreciate the cultural diversity and beauty of the world. The data collection process also showcased a clear evidence of mimetic impulse, which is why most textile patterns around the world are inspired by nature. We also discovered firsthand that AI does a good job at clustering visually similar textiles, which also raises an interesting question of how AI will view cultural differences and similarities in the future.