Mirasys VCA Deep Learning

Requirements

• Nvidia GPU with CUDA cores

  • A NVIDIA GPU with CUDA Compute Capability 7.5 or higher

  • Depending on GPU CUDA cores, how many Deep Learning channels you can use on the system

• The latest NVIDIA graphics drivers (at least 460.73 or higher).

• CUDA Toolkit

• Mirasys VMS 9.4 or newer

• Deep Learning object files

Installation

1. Install latest Nvidia drivers to the system

2. Download Mirasys VCA Deep Learning package from Mirasys Extranet

3. Extract the package

4. Browse to folder CUDA Toolkit

5. Install CUDA Toolkit with all features

   1. You can find detailed installation guide here.

   2. Some features are not installed because Microsoft Visual Studio is not needed to install but the toolkit is providing example files

   3. If you have installed already Mirasys VMS, before copying files VMS services need to stop

6. Stop services: WDServer, DVRServer and SMServer

   1. This is not needed to do if you are using V9.6 or newer

7. Copy the content of the VCA Deep Learning files folder to C:\Program Files\DVMS\DVR\vca\bin location

   1. This is not needed to do if you are using V9.6 or newer

This path is the default installation location of Mirasys VMS
If you have installed Mirasys VMS to another location, copy files there

1. Start WDServer, DVRServer and SMServer services

   1. This is not needed to do if you are using V9.6 or newer

Now you have installed and are ready to go with Deep Learning tracking.

Licensing is done via local VCA Deep Learning licensing or using License Server (Virtual Environment or if you want to handle licenses in one place).

Some cases detection may not work correctly. Please try to increase image quality or move/zoom camera image to closer wanted detection area.

Models are trained using clear images and some cases when using black/white image or thermal camera image this may cause that detection is working correctly. For this you can try use Deep Learning Filter with Object Tracker.