Face analysis application

Project background

Photographers usually take more than a thousand pictures at an event, party, or at festive celebrations. It’s very challenging to find the needed photos among so many files.  

Our client, a Japanese company, provides photo services at schools, kindergartens, camps, and various children’s events. Our client was in search of a professional face recognition app development team.   

The system has to work with such complex cases as detection and recognition of small, rotated, collapsed faces, as well as faces detection in group photos.

The general app workflow: 

– Photographers conduct photo sessions and upload photos to the system.

– Parents log in to select 1-2 photos with their children.

– Upon a request from the parents, the system returns the results of all found and recognized faces.

Business needs

The client’s staff photographers usually take more than a thousand pictures at an event, and it’s challenging for parents to find their kid’s photos among so many files.  

– Our client was looking for the best face recognition software development team and ways to improve face recognition accuracy of their solution for picking the right child’s photos.

– Our client needed to improve both solution efficiency and face recognition accuracy.  

Product features

1. Face detection
2. Face recognition of not-frontal faces
3. Face recognition of overlapped faces
4. Small faces recognition
5. Children with emotions recognition
6. Face (and image) is turned in the roll direction
7. Image in more than Full HD resolution
8. Face captures big area on the image

Solution

To begin with, the CHI Software team conducted a study to find a service for improving the accuracy of face recognition in photographs.

We conducted a comparative review of Azure, Amazon, and Kairos APIs to figure out the best way to address the following issues:

– Full face recognition

– The number of people in the photo

– The size of the face on the photodetector

We evaluated a custom software solution to cover all the above-mentioned issues. After that, our team proceeded to the face recognition app development phase.

To increase solution efficiency and face recognition accuracy, we did the following:  

– We set up a cloud environment for development and product hosting. We recommended AWS EC2 as an infrastructure platform, AWS S3 storage, and RDS database as server configuration model;  

– We also performed system development and customization. We improved the accuracy of rolled and rotated face recognition (by 45-90 degrees), background recognition, extremely small or big face detection.  

– We provided opportunities to automate monitoring, analytical and comparative manual processes.  

  After system integration, we are now providing support services.

Our technology stack

• Amazon
• Kairos
• Azure
• OpenCV
• Docker

Key achievements delivered

1. CHI Software researched and provided detailed documentation on how to make a face recognition software that meets our client’s needs.
2. We designed and developed a face detection and recognition system allowing photographers to upload numerous photos to the cloud.
3. Thanks to face recognition technology, parents can find their children’s pictures and get all the photos much faster.
4. The solution saves photographers and parents time and improves the client’s existing workflow.