New Experience is Amazing: Face Detection App Withing 24 Hours

At v-jet group we respect everyone’s efforts. Company’s success is the result of dedicated teamwork. That’s why recently we decided to arrange local hackathon and create masterpiece instead of just working app. Our team was impressed with the growing popularity of face detection apps. So, we gathered team of developers for local competition priced with paid vacation for all participants. Thanks to combination with creativity our developers made their best to bring unique and outstanding vision of the product.

Technologies we used.

After some time spent on research we figured out that none of the suggested free ready-made solutions performs well (we found 8). Every found lib should be improved and/or updated to your needs. You will never know how new experience will influence on your workflow, that’s why our team decided not to test the water and use reliable and well-known tech stack for new experience.    

Server: CentOS 7.x., PHP 7.x., MySQL, Python

Backend: We have been looking for ways to detect faces in photos with PHP. Among possible solutions for doing it with PHP, we pointed out the OpenCV opensource lib that was originally developed by Intel.

CONSIDERATIONS

• Complex, hard to use software designed for computer vision professionals only.
• Covers a wide area related to computer vision.
• Limited depth in relation to face detection / recognition.
• SDK only.

For hackathon using OpenCV opensource lib is something like training wheels, because for commercial applications it has poor recognition rates.

Frontend: Angular1 + Yii2

The process we created our application and how our technology fit into it?

We created our software to use the camera, then use OpenCV to try to detect faces in images in real time. OpenCV already contains many pre-trained classifiers for face (4 parameters), eyes, lips, ears and nose. Those XML files are stored in openCV/data/Haar Feature-based Cascade Classifiers/ folder.Once a face was detected all info about how do they match to each other appears on the screen.

What challenges did we face while building the application? Any success?

The biggest success we had that everyone was engaged into a process and made it’s own input into development process. Our team implemented face detection from the camera only. Sure we wanted to extend the application and make it more sophisticated: combine the id with the name, then show the confidence of the prediction, recognize the emotion... and and and. Nevertheless,  this was more than enough for local hackathon competition to assure that we are able to handle such kind of apps in future.

What’s next?

So,that was just a simple dive into building face detection app withing 24 hours.For further improvements of face detection to implement it as the core feature for other apps and websites, we’re going to spread our expertise and try other libraries as well as technologies. Powerful combination of Machine Learning/AI and the Cloud are a match made to follow the future improvements. Very soon this combination will be applied to videos. Something like Facebook tagging for Video or YouTube detecting product placement, but much more complicated. It is something we should think over and bring new idea of hackathon to live.

I believe we will bring the idea further and use such kind of technology to detect someone's move. Who knows, maybe we'll find out how to detect someone stealing in a shop, or needing help to get to their car – the future of Computer Vision, AI and Neural Networks is exciting!