Huematch
Huematch utilizes Intel OpenVINO™, OpenCV, and other technologies to bring a new form of ease and accessibility to the beauty industry. Huematch uses AI to perform an analysis accurately predicting the perfect make-up shade and foundation that "matches" the "hue" of your skin! ...learn more
Project status: Under Development
oneAPI, Internet of Things, Artificial Intelligence
Intel Technologies
OpenVINO
Overview / Usage
We started Huematch out of necessity. With COVID-19 mandating quarantine, it became impossible to go to a mall or a beauty store to try out different foundations. And for BIPOC, it’s even more difficult; rarely do brands have comprehensive shade ranges.
All people deserve a foundation that fits their unique hue. No brand biases here. And their hue should be accessible from anywhere - instantly. It’s the digital age, after all. So Huematch was born. Built with cutting edge technology for a near perfect match. Huematch helps you discover your true colors.
Methodology / Approach
Huematch uses Intel OpenVINO™’s face detection capabilities to map the key areas of the face from an uploaded image. It then uses the pre-trained facial-landmarks-35-adas-0002 model in order to detect relevant landmarks of the face that will serve as the boundaries for the color detection region. Landmarks 6 and 20 mark the edge of the right nostril and right cheek, respectively. We use the coordinates of these landmarks to determine a facial region – a rectangle in the center of the right cheek – that will be used to detect the user’s average hue. This particular region of the face is the most popular when testing for foundation shade matches, and therefore where we will match our user’s hue to a foundation shade.
The project then uses OpenCV to create a mask over a cropped image of the facial region of the cheek. OpenCV is then manipulated to detect the average hue of this mask, giving us the Red, Green, Blue (RGB) values as well as the hex value of the hue.
We then check if the RGB values of the average hue of the user are within a particular range of the RGB values of any of the foundations in our database. If the RGB values fall within a particular range of a foundation, the user’s “hue” is “matched” to that foundation and the process is complete! You’ve discovered your true colors.
In future versions of the application, we will implement additional facial regions (including the left cheek, chin, nose, and forehead) and take the average hue of those areas for a most accurate hue match. As the technology picks up traction, more brand partnerships means more match options, as users can elect the formula they desire and specify skin conditions that warrant a particular product.
Technologies Used
AI/Deep Learning: OpenVINO™ 2020.4
Graphics/Visualization: OpenCV
OS: Ubuntu 18.04.2 LTS
Software: C++
Hardware: Intel NUC Kit NUC7i7DNEH
Documents and Presentations
Other links
Collaborators
There are no people to show.