News 10 Jun 2022
How our leading synthetic data enable retail to be more sustainable
Indivd has built the world's largest and most reliable database for synthetic data. This accomplishment has significantly increased the accuracy and reliability of our people counter, which helps stores and malls to improve their sustainability.
- Our synthetic database contains over a billion images of virtual people in different clothes and different virtual store environments, says Leonard Johard, Lead Scientist at Indivd.
Indivd has, for some time now, used synthetic data and synthetic people to train our models and algorithms, in our work to help stores and malls become more sustainable.
- A big issue when it comes to increasing the stores' sustainability is that they must be able to understand the behavior of their visitors. Visitors' behavior needs to be the basis for which products the physical retail design, produces, and distribute. Without that understanding, physical retail will continue with inaccurate production and transportation. In order to be able to build these insights, underlying data is required, which leads to major GDPR problems. This is an important background for why we have invested time and resources in building the world's largest and most accurate synthetic training data for our people counter, says Leonard Johard, Lead Scientist at Indivd.
We are improving the accuracy of our people counter with the support of leading experts
Indivd's RnD team collaborated with 3D modeling experts in this project from the global game development industry, the film production industry and the construction industry. The experts also have experience in developing virtual urban environments and museums.
- Our ambition was to use synthetic data to further improve the accuracy of our people counter. This is something we are proud to be able to say that we have achieved, says Leonard Johard.
The database contains over 6 million different virtual people in different store and mall environments
We have, with the help of synthetic data, managed to create virtual store environments with virtual visitors. This has made our database the world's largest database of realistic visitors in realistic clothing.
- We have created over 6 million different virtual "people" in different body profiles and different situations. In total, we have more than 1 billion different images, says Leonard Johard and continues:
- A special challenge for us has been to produce various garments with random color schemes adapted to different body shapes. We have developed this together with several training methods, where different "people" interact in different store environments. In addition, we are constantly introducing new garments. This database allows us to take as many photos as we want and place them in a realistic store or mall environment with differentiated people counters, angles, and qualities.
We have developed a completely new generation of privacy-friendly people counters
The demand for synthetic data is now increasing from both companies and researchers.
- With this type of data, our people counter is far ahead of everyone else. Of course, we have great use of our database to develop our privacy-friendly people counting, but also help the physical retail to become more sustainable, says Leonard Johard.
The next step, which has already begun, will be to supplement the synthetic still images with a video database. The virtual visitors will then be able to move around to perform selected specific behaviors in the various virtual stores.
- Our new graphics engine will increase the accuracy of our people counter further. This will at the same time increase the stores' and malls' ability to improve sustainability, while we can reduce the incidence of inaccurate design and production as well as unnecessary shipping, Leonard Johard concludes.