Analysis of video streams using neural networks and real-time computer vision

In the world of video streaming, high load and real processing time have long been standard conditions. But this applies to conventionally "classical" tasks. With heavyweight neural networks in computer vision the story is different: few can handle simultaneously dozens of video streams on one server and process them on the fly.

From the talk you will learn why such problems appeared in the first place, how the speaker's team solved them and why they left Python as the main language of our solutions.