Real-time video analysis for the web
For multiple use cases like face filters, object detection and tracking, emotion detection...

Object detection in the browser

We have developed a deep learning engine able to analyze a video stream in real-time in the web browser. It brings advanced computer vision to web applications. Then augmented reality becomes possible in a web context.

We have created a framework around our deep learning engine to generate training samples, instantiate neural networks, train them, run neuron networks and build applications. It is a highly optimized end-to-end system.

Write once, run everywhere

Deep learning based computer vision

Our computer vision framework only relies on open and standardized technologies like WebGL or WebRTC. Thus it works everywhere: on websites (in the web browser), on mobile applications (Progressive Web Applications), on desktop applications (with Electron) and even on embedded hardware (using Nvidia Jetson).

We are flexible

We can create and train outstanding artificial neural networks to solve any kind of real-time computer vision problems.

Our first demonstration is a glasses virtual tryon web application. You can check it on jeeliz.com/sunglasses. It was released at the beginning of 2016 and it was the first web application implementing deep learning in the browser matching a commercial use-case.

In this example, the neural network inputs an image and outputs whether it is a face or not, what are the position and the rotation of the face, and even lighting parameters. Then the glasses 3D model is rendered over the video at the right place and orientation and coherently enlighted. This process is repeated dozens of times per second, more than once per new video frame. Thus the virtual glasses follow the head smoothly and accurately.

We bet on the GPU

Our deep learning engine runs on the graphic processing unit (GPU). Nowadays even the cheapest mobile phones have a dedicated GPU. The GPU is better than the CPU for parallel computing. GPUs are less impacted by the Moore’s law than CPU’s because they can scale horizontally. So their computing power is still increasing consequently.

What makes the difference?
Speed matters

We are often asked: Why shall I a use Jeeliz librairies since tensorflow.js seems to do the same job? tensorflow.js also runs in the browser in JavaScript/WebGL, the developer community is huge and it is made by Google…

Tensorflow.js is great. But this framework was built with a top-down approach: JavaScript/WebGL is just another way to train and run deep learning models in the browser.


At the opposite, we started from what we get in a web development context: the JavaScript/WebGL workflow. Then we contraint the neural network structure in order to apply advanced optimizations.  Thanks to this bottom-up approach, our deep learning engine is on average 5 times faster than Tensorflow.js. It makes our technology fast enough for real-time video processing.

Read more: Why is our technology unique?