Blog #1: Solving the Fashion Fit Crisis with Open Source Tech

For my field study, I am trying to develop a digital virtual fitting room to address the "trillion-dollar problem" of online returns, which can soar to 40% because of poor garment fit. While I initially considered using a proprietary engine, I have pivoted to a more specialised, open-source pipeline that prioritises anatomy accuracy and real-time physics with less programming and extensive customisation.

One of the key findings of my research is that many professional fitting tools tend to produce avatars that are “too smooth” or perfectly symmetrical. Such models do not take into account the realities of age, including the distribution of body fat, contours of muscle, sagging and asymmetry. To fill this gap, I am using MakeHuman for parametric modelling and MeshLab to clean and process 3D point clouds to create a more realistic human form.

I rigged the characters with "Mixamo" and simulated the cloth using an enhanced PBD framework. The cloth simulation is believable and optimised to be used in normal devices. This student friendly approach aims for hyper-personalised sizing with an error rate of less than 10%, dramatically improving consumer experience and cutting down on industry waste.