Challenge, approach, and impact

• Algorithm development

  The main challenge was also our core task. We had to develop Machine Learning models for an unprecedented use case. We’ve experimented with many techniques to process 3D scans of a human body to get the needed information.

• Multiple vendors

  One of the biggest management challenges was connected with a multiple-vendor situation. One team worked on mobile application development, and another team provided 3D scanning technology. We had to set up a practical communication framework to process scans and successfully integrate our solution with the application.

• 3D scan processing

  Some of the scans we worked with were low-quality (e.g., obtained using a mobile phone). Still, our ML system had to be able to process them effectively, as they were a big part of the use case.

• Insufficient training dataset

  A lot of initial challenges were connected with a low volume of quality scans we could use to train our Deep Learning models. Still, we couldn't just wait for more data but had to actively work on a solution.