Skip to content

Introduction

Camera recommendations

Network camera

It is recommended to use wired network cameras connected with an Ethernet cable for the best performance. Wifi cameras can work, but will almost always loose connection or fail at some point. They are not as reliable as wired cameras.

Camera manufacturers

Cameras from the following manufacturers are known to work well with Yawcam Ai:

  • Hikvision
  • Dahua
  • Axis

I have had several issues with Reolink cameras, so I would not recommend them.

Video format

  • H.264
    Always use H.264 for video encoding when possible. This will maximize the compatibility and also reduce the CPU load on the server.
  • H.265
    Even though H.265 is more space efficient than H.264, it is not as widely supported and will require more CPU power from the server.
  • Substream
    If your camera supports a substream, use this for improved performance. The substream is a lower resolution stream that is used for motion detection and streaming. The main stream with higher resolution is used for recording.

GPU info

It is highly recommended to run the object detection on a GPU instead of the CPU 1. A graphics card will increase the speed considerably. To run the detection on a GPU you need an Nvidia Card with CUDA support using the Maxwell architecture or later 2. You also need the special CUDA drivers for the graphics card.


  1. The detection will work on the CPU as well, but it will be much slower. 

  2. Yawcam Ai is using the CUDA drivers from SDK version 12.x. This means that the GPU needs to support CUDA Compute Capability 5.0. This is the case for Nvidia GPUs with the Maxwell architecture or later.
    Please refer to the CUDA GPUs page for a list of supported GPUs.