Design and Development of an Application for the Generation of Garment Patterns Based on Body Measurements Using CNN

Geraldine Curipaco, Jeiel Tarazona, Daniel Subauste

In recent years, the growing consumption of products and services over the Internet has led companies to take their business models to virtual platforms, so retail fashion companies, such as ateliers, need to adapt to new market trends. Ateliers specialize in making garments to the customer's measurements, so this process requires a high level of time, cost and personnel specialized in taking body measurements and pattern making. There are several technological solutions to obtain body measurements and other solutions to obtain garment patterns, however there are no solutions that integrate both processes. Therefore, we propose an application to produce dress garment patterns tailored to a person from photos, using image processing and convolutional neural networks (CNN). Our proposal starts by obtaining a frontal and a lateral photo of the person, as well as her height, and then processing the images to obtain the body measurements by means of a CNN. Finally, we proceed to adjust the patterns of the garment required with the measurements obtained in order to give as a final result the dimensions of the dress's garment patterns.

Full Text