End to End Computer Vision in Elixir: Making it work in Production

Walid Salah


The task was to replace a proprietary video recording platform with an elixir based solution that had to perform in some very harsh physical environment with unreliable power and bandwidth. Evercam chose to implement a membrane.stream-based solution deployed using Nerves Project and leveraging Livebook and the emerging elixir machine learning & computer vision ecosystem. What we found was that it was relatively easy to put something together very quickly, but to actually make it work in production meant solving a lot of edge case problems. The resulting project, Ex_NVR has been made open source and is intended to be a kickstart for successful computer vision projects deployed on the edge.

Key takeaways:

  • Priorities to consider when trying to solve for real world constraints.
  • Developing for embedded devices.
  • Computer Vision in Elixir (AI & Traditional).


  • Accelerate other people’s computer vision projects.
  • Explain some of the pitfalls that ExNVR tries to address.

Level: Introductory and overview

Tags: Video, AI, Computer Vision

Format: Virtual