HyperML: Deep Learning based image processing on mobile

Hyperconnect is researching various technolgies to effectively implement data processing through machine learning on mobile devices, servers and browsers. HyperML is a platform based on deep learning-based image processing, and consists of a mobile-optimized inference engine and pipelines for data processing and training.

*Real-time computer vision solutions based on pre-trained models will be available as SDKs on Android / iOS platforms in the near future.

Mobile-optimized deep learning engine

The HyperML inference engine runs on mobile devices. Optimized for the mobile environment, which is typically characterized by limited memory and processing power, the technology ensures real-time results without the data ever being processed by the server.

Artificial intelligence-based image recognition

We have developed a deep learning-based, mobile artificial intelligence system which can, among other things, recognize the current environment wherein the user is situated or filter out inappropriate imagery. This technology is already being used in Azar to provide a more satisfactory matching experience for the users.


HyperCut is a real-time image segmentation technology that processes real-time video feed, such as video frames from camera, and separates people from background. Trained on rich data, this technology is very precise and can separate even the most complex backgrounds.

Other Techs

Real-Time Communication

WebRTC technologies, optimized for the mobile environment, implements peer-to-peer and multi-party connections as well as broadcast services.

Learn more

Mobile AR Graphics

Based on computer vision and machine learning technology, Hyperconnect is capable of visualizing graphic effects of AR (augmented reality) on mobile platform.. This technology is used to change the background of the person in the video, recognize facial features in 3D in real-time.

Learn more

Ready to build with Hyperconnect?

Create new possibilities through the technology of Hyperconnect.