Digital blendshape for intimate apparel design

Partner: HYPERcurve

Motivation

Currently the lingerie sector remains one of the most traditional fields within the fashion industry.  From the creation process (which can take up to 18 months from sketch to delivery to client), to the fitting & shopping experience. Designers do not have the right tools to incorporate breast shapes and fit in the design process. Samples can only be assessed after creation and are usually fitted on live models. By implementing innovations such as 3D prototyping, the design/development and production processes can be optimized (see figure below). 3D digital sampling is faster, cost effective and is more sustainable as it eliminates the use of physical materials.

Development of a lingerie collection

Problem definition

In the past few years, some 3D human models have become available. However, these models are often based on 3D scans where women wear a sports bra, neglecting the effect of an underwire and brassiere model type (see figure). The research question for this project is defined as:

“how can we best integrate the diversity of brassiere models, shapes and sizes into a realistic personalizable 3D body model for the lingerie industry?”

A sub-research question is

“how can we accurately generate a large range of (in-between) shapes and sizes from a single digital mannequin?”.

The goal of the project is to design a digital tool that can be used by many lingerie designers, pattern makers and manufacturers that will help to speed up the internal  processes while making it more inclusive for shape and size variations within the user group. This tool should be able to:

  • Be used in software like CLO3D
  • Be adjustable in realistic shapes and sizes
  • Predict the effect of different standard brassiere models (eg. balcony, deep plunge, etc)

Besides the obvious benefits of replacing physical sampling in the visual design process, there are other possible benefits in the design process with a 3D design tool. A 3D human body mannequin can be build using statistical shape modeling, a methodology to map the variations into one 3D avatar. This avatar would allow for a designer to adjust its design to the variations in women’s breast volume, shape, skin tone and personal preference. This opportunity greatly helps to solve the inability to fit and visualize on all types of bodies within all sorts of lingerie/bodywear brands. In a final stage of this tool and its methodology we can imagine it would solve the issue of inclusivity of shape and size. Firstly, it can purely solve the need for visualization of all sizes while shopping online. Later it could be the base for a completely new way of designing and producing such as mass customization. for many women within the user group, it is hard to find the right size and design, especially for those with either large or small sizes. A tool that includes all data on variation rather than standard sizes could shift the development process completely.

Activities and task division

Scanning

In the preliminary research by HYPERcurve, relevant research parameters, such as brassiere models and sizing systems, were investigated. After that, a two-fold research was prepared and executed by HYPERcurve and the Fieldlab UPPS researchers. Participants were recruited within the customer network of the company HYPERcurve. Study 1 included two female subjects, testing 20 brassiere models. Study 2 included 19 female subjects, testing three different brassiere models (of which 2 constant and 1 ‘perfect fit’ as judged by customer and lingerie expert).

Brassiere models

Beforehand, data about bra size, body measurements, body weight, age and breast firmness were collected. Breast firmness was measured using silicone sample samples for comparison, in which participants compared their own breast firmness to the silicone samples with different shore values.

Silicone samples of breast firmness comparison
Scanning of a participant with Artec EVA structured light scanning

Blendshape tool creation

The Fieldlab researchers processed the scans and created a statistical shape model, to be used as a plugin for the software Blender (see image). The tool allows a user to adjust body properties (underbust and overbust girth), bra properties (cup and band size) and bra fit properties (based on the two bra models). However, this variable can be set on all values in-between those two fit standards, to generate a whole range of bra fit results, releasing the connection to a standard bra model. Using the tool, different meshes can be exported to be used for many purposes.

Lingerie mannequin blendshape tool

Results and conclusions

To validate the tool, several heatmap analyses were executed between two meshes using Haussdorf distance filters. The illustration below shows a heatmap visualization illustrating difference between a raw body scan of a subject with bra size 75D and a simulation generated with the tool, using the body properties known about the subject. The red color indicates proximity, where blue indicates larger distance. Visible is that the area of the breasts shows good proximity. Larger distances around the belly can be explained by the fact that those body properties are not used for the creation of the tool and that they are independent from breast shape and size.

Heatmap of Hausdorff distance between a raw scan and a simulation by the blendshape tool with similar breast and bra properties

The figure below shows a similar illustration, but now two manikin simulations of bra 2 and 3 are compared. On the left illustration, the breast texture is set to very soft, whereas in the right illustration, this setting is set to very firm. The difference between a very soft and very firm manikin can be seen in the image with side view below. Visible is that for soft breast types, the largest difference in bra fit is located more on top of the breast, whereas for firm breast types, this difference is located lower on the breast.

Heatmap visualisation showing Hausdorff distance between the Prima Donna bra and the Wacoal bra for very soft (l) and very firm (r) breast types
Very soft breast shape (l) and very firm breast shape (r)

Future use

Although the tool is versatile and detailed as it is, it is still only based on a limited set of subjects (19) and limited variation in brassiere models (2). This tool should be optimised by adding more subjects, with more size variations and wearing more different brassiere models. The different brassiere models could then also be classified using technical properties, as illustrated in the figure below. With further enhancements the tool could become practically useful from end to end, meaning; from Design to marketing to sales processes. One of such enhancements would be that the tool visualizes full body mannequins. Linked to existing full body anthropometric data. Another important visual aspect would be the nipple visualization and realistic skin textures.

Technical properties of brassiere models to be used as blendshape variables

In future versions, the tool would potentially include soft tissue parameters and additionally have the option to export pattern and product specifications based on the created body shape result. This way the tool would supporting a designer both visually and technically by clearly indicating the possibilities and limitations within the product development the ‘field of play’.