“3D Custom-fit” is the term we created to refer to the technologies we are using during the mass-customization process. For your better understanding, we divide it into several parts:
3D Scanning and 3D Model Analysis Algorithms
The 3D scanning module contains three parts, the laser gird projector, the receiver, and the RGB camera. They produce point cloud data and the RGB video stream. When you hold the scanner and move it around the object, the scanner records multiple frames of these data at different locations and angles. And SLAM (simultaneous localization and mapping) algorithm is used to stitch all the frames together, giving us a 3D mesh model with texture.
Then we need to perform the face landmarks detection on the 3D mesh model so that the computer would understand where are the eyes and where is the nose etc. This will provide enough information for the parameter adjustment algorithm to control the parametric modeling algorithms. The are many options for the face landmarks detection, the most effective one is using deep learning, which is quite mature nowadays.
Parametric Modeling Algorithms
Parametric modeling algorithms refer to the algorithms that take several parameters as input and generate the complete 3D model of the product as the output. By changing the parameters, the product design will change accordingly. Building the parametric modeling algorithms is the most important part of the 3D Custom-fit workflow because the algorithm is the product itself that determines all the designs and specs of the product.
This concept of parametric modeling is very confusing because some other CAD software has occupied this phrase. However, the difference between them is huge. Our parametric modeling algorithms are designed to be adjusted primarily by computer, while the old one refers to CAD software that needs to be operated by engineers when changing the parameter values. In a nutshell, when we create parametric modeling algorithms, we focus on how to automate the design process, reducing human interference and letting the computers do the job.
Currently, we are developing these parametric modeling algorithms on the Grasshopper platform, which has been widely used by architects for programmatic modeling and generative design. Other CAD software companies begin to publish similar products like Grasshopper in recent years. We think this is a good trend because competition means improvement.
As a new way of making things, 3D printing offers much more freedom in the design, and it is also influencing the business model. Designers and engineers need to master new tools like the Grasshopper to build 3D Custom-fit projects. We hope, in the future, Skelmet could bring more attention to this field and more talented people could join us.
Parameter Adjustment Algorithms
After we create the parametric modeling algorithms, we need to give them the correct parameters so that the custom-designed product would fit perfectly for the user. This means we need to establish a mapping relation between the 3D scan and the parameters.
We solve this problem with three methods: 1. Let engineers manually set the right parameters according to their knowledge and experience. 2. Write simple and robust mapping functions without human interference. 3. Use end-to-end machine learning when having enough training sample sets. In practice, these methods are combined together to achieve an ideal fit for the user.
Finally, every product is different in terms of parametric modeling algorithms, and the same is true with the parameter adjustment algorithms. And the more orders we accumulated the more accurate the parameter adjustment algorithms will become. So we’ll constantly refine the parameter adjustment algorithms and collect feedback from users.
Order and Manufacturing Management System
From our experience, the order and manufacturing management system are also critical to build a successful mass-customization business. Because when hundreds of orders are being processed, you have to track the production process of every part so that they don’t turn into a mess.