3D head reconstruction solution
Our client is a creative design agency that provides branding, advertising, graphic design, and virtual modeling services for individuals and corporate customers. The agency has its own in-house souvenir production.
Computer Vision solution for a cosmetic retailer
This AI-powered mobile solution is created for beauty retailers. It personalizes the user's skincare routine to make anti-aging treatment a smarter and more customized experience.
Project background
Not so long ago the sale of cosmetic products depended mostly on the physical store experience: the customer needed to try out different shades or beauty products before they could find the right one that suits them best.
Retailers now have to meet the growing expectations of the customers for engaging experiences in stores. That’s why the consumers’ experience in beauty stores is now greatly transformed by technology, namely by diverse mobile opportunities: starting with chatbots and finishing with AI-powered solutions to pick up needed products fast.
Our client, a cosmetic retailer, has a variety of beauty items, including products for all skin types, under-eye care, and decorative cosmetics. They were looking for innovative mobile solutions that can analyze the customer’s face and type of skin, and then offer the best-suited product and, thus, improve their customers’ experience.
- November 2020 – April 2021
- New York, US
- Beauty/Health/Retail
- C++
- Open CV
- Dlib
- Caffe
- Swift 4.2
- Alamofire
- REST
- AVFoundation
- CoreGraphics
- CoreImage
- CoreAnimation
- Multithreading
- UserDefaults
- Firebase analytics
- SafariServices (API)
- PHP 7.2.12
- MySQL
- MariaDB (AI)
- AWS EC2, S3, RDS
- Face analysis Analyzes collected user data received from a photo (selfie)
- Custom recommendations Based on the collected data, makes recommendations on available beauty products
- Helpful hints Offers hints and tips for the buyers to simplify the selection of items in the catalog
Business needs
The client’s beauty company has a wide range of beauty products and many customers in the United States. To serve everyone, the company needed to train a large staff of consultants. Therefore, to optimize processes and minimize human labor, an idea of an app appeared. This solution should help end customers choose the right cosmetics themselves. Thanks to AI-powered mobile solution the number of dissatisfied customers should decrease, and the costs of hiring and training new consultants should be optimized.
Solution
CHI Software developed a mobile application based on Computer Vision, and photo face recognition. Here’s how it works:
-
A user takes a selfie
-
Using the detectors based on the histograms of forfeited gradients from the Dlib library, the program can accurately determine the user’s face on the photo
-
Using the same approach, we were able to train a system that can detect 126 feature points on the user’s face
-
The app then analyzes the unique facial features of a user and adopts the user profile
-
The applied AI technologies then help select the needed skincare
-
Our system allows to perform a detailed/smart analysis of the user’s face
-
The app now provides personalized skincare and beauty recommendations
Client values
- The client’s new mobile-based beauty app helps users not only to navigate better in the range of items but also to buy cosmetics according to skin type and preferences
- The selection of cosmetic products has become a simple and convenient automated process that increases consumers’ loyalty
- CHI Software researched the marker leaders and popular beauty tech solutions
- Designed a prototype to demonstrate how this solution can help the users
- Developed AI-powered mobile apps for iOS and Android users
Employee testimonial
The beauty app was an interesting project in terms of technology with a revolutionary idea to improve the beauty industry behind it. I've practiced and boosted my skills in implementing Computer Vision in Face Analysis business solutions. As for challenges, the most difficult part was the detection of skin defects like acne. There needed to be clear criteria for their detection since defects' visibility can be very different under different lighting conditions.
Artem Luban C++ Developer