Videos are powerful communications tools, and now even more so after the world moved to online space adjusting to the COVID-19 pandemic. However, most professionals find it difficult to create a good-quality video. Remote work means working from home or one’s sleeping room neither of which are ideal places to record a professional video. Currently, there is no solution that allows for the easy addition of high quality virtual backgrounds –usually, expensive green screens and post-production are required. Using a virtual background in live video platforms (e.g. Zoom), has its shortfalls too, as the speaker cannot always be distinguished from the background, challenging hair styles, furniture is not properly identified, etc…, creating a blurred boundary and unrealistic result.
Maekersuite is developing novel AI solutions to solve this challenge and they have received ELISE funding to support them in this endeavor. Stephanie Demetriou, Maekersuite co-founder explains that “the ELISE grant is to help us develop and deploy HD, real-time, virtual background tech on embedded devices (mobile phones) so that anyone can record wherever they are.” The project allows anyone to record video wherever they are and enables professionals that are not video experts to add High-Definition virtual backgrounds and create professional content. “We hope to develop this technology to also be used on devices with lower quality cameras (e.g. desktops) which grows the market even further” Demetriou added.
Furthermore, Deep Learning Computer Vision algorithms have seen an exponential growth in usage and applications. However, when working with images/videos containing people, not enough attention has been paid on how sub-groups can be affected by these algorithms and despite this growth, datasets are still known to be biased against minority groups, for example, according to a study by the Gender Shades Project, models identifying darker females are significantly less accurate than that of lighter males .
At Maekersuite, we strive to cater to a global audience, deploying products and features where this kind of mistakes do not occur. To achieve this, we have generated our own synthetic datasets and continuously train, improve and test our models against the AGE criteria (age, gender, ethnicity).
Maekersuite was founded by Stephanie Demetriou and Philip Werner and embarked on a mission to help professionals become better communicators through video. They currently have about six employees and are working in the Technology and Media space.
Interested to learn more? Check out the Maekersuite website here.