Where can I install video analytic plugins?

/piattaforma.jpg

Available Technologies

Plugins are multi-platform and can be provided in three different options:

  • Edge side: directly on board of specific models of cameras
  • Server side: compatible with any IP camera
  • Embedded: as an integrated solution hardware + software

Specifically, the following is a map of compatibility for each platform:

compatibility_map

Edge Version

This option is suggested for customers equipped with cameras that can host the software on board and who want to reduce network load. Plugins operate on board of specific models of cameras and only events of interest (e.g.: the passing of a person through a gate) are sent to external systems. Elaboration is processed on cameras with a resolution up to 4CIF.

N.B.: this option is available only on specific models of these brands:

Server Version

This option is suggested for customers using IP cameras that do not allow the on-board installation of the software on the camera; thus, it is required an external server with the plugins. Video streams can be acquired through the RTSP / ONVIF standard protocols. The server can process simultaneously more streams on the same machine with a higher resolution and frame rate than cameras, notifying events of interest to external servers. The specifications needed to analyze video streams can vary, depending on the specific plugin.

Traditional algorithms

For the plugins using traditional machine learning approaches (not belonging to DEEP VISION products line), supposing to work with a maximum frame rate of 30 FPS, it is required:

For 1CIF resolution:

  • RAM: 2GB (for the operating system) + 50 MB for each plugin; for instance, 8 plugins will require a total of 2,4 GB
  • CPU: Intel i7 or XEON, 1 core (or Virtual Core) each 4 plugins; for instance, 8 plugins will require 2 cores (or on a virtual server based on an XEON processor with 8 plugins, 2 virtual cores are required)

For 4CIF resolution:

  • RAM: 2GB (for the operating system) + 100 MB for each plugin; for instance, 8 plugins will require a total of 2,8 GB
  • CPU: Intel i7 or XEON, 1 core (or Virtual Core) each 2 plugins; for instance, 8 plugins will require 4 cores (or on a virtual server based on an XEON processor with 8 plugins, 4 virtual cores are required)

Algorithms based on DEEP-LEARNING (no GPU required)

For the plugins that use deep learning based algorithms, but without requiring GPU, different specifications are required. In fact, for the correct use of the AI-FIRE-DEEP, AI-SMOKE-DEEP, AI-FIRE+-DEEP, AI-SPILL-DEEP and AI-PARKING-DEEP plugins you need:

For a 1 CIF resolution, assuming that it works with a maximum of 16 FPS:

  • RAM: 2GB (for the operating system) + 100 MB for each plugin; for instance, 8 plugins will require a total of 2,8 GB
  • CPU: Intel i7 or XEON, 1 core (or Virtual Core) each 1 plugins; for instance, 8 plugins will require 8 cores (or on a virtual server based on an XEON processor with 8 plugins, 8 virtual cores are required)

For a 4 CIF resolution, assuming that it works with a maximum of 4 FPS:

  • RAM: 2GB (for the operating system) + 200 MB for each plugin; for instance, 8 plugins will require a total of 3,6 GB
  • CPU: Intel i7 or XEON, 1 core (or Virtual Core) each 1 plugins; for instance, 8 plugins will require 8 cores (or on a virtual server based on an XEON processor with 8 plugins, 8 virtual cores are required)

For a 1080p resolution, assuming that it works with a maximum of 1 FPS:

  • RAM: 2GB (for the operating system) + 500 MB for each plugin; for instance, 8 plugins will require a total of 6 GB
  • CPU: Intel i7 or XEON, 1 core (or Virtual Core) each 1 plugins; for instance, 8 plugins will require 8 cores (or on a virtual server based on an XEON processor with 8 plugins, 8 virtual cores are required)

Vice versa, the plugin AI-BIO-DEEP and AI-FACEDETECT-DEEP requires, supposing a frame rate of 20 FPS:

For a 4 CIF resolution:

  • RAM: 2GB (for the operating system) + 2 GB for each plugin; for instance, 8 plugins will require a total of 18 GB
  • CPU: Intel i7 or XEON, 1 core each 1 plugins; for instance, 8 plugins will require 8 cores.

The plugin AI-CROWD-DEEP requires, supposing a frame rate of 5 FPS:

For a 4 CIF resolution:

  • RAM: 2GB (for the operating system) + 2 GB for each plugin; for instance, 8 plugins will require a total of 18 GB
  • CPU: Intel i7 or XEON, 1 core (or 2 Virtual Core) each 1 plugins; for instance, 8 plugins will require 8 cores (or on a virtual server based on an XEON processor with 8 plugins, 16 virtual cores are required)

Algorithms based on DEEP-LEARNING (requiring mandatory GPU!)

Finally, for plugins that instead use an approach based on deep learning algorithms and also need the GPU (NVIDIA compatible), different specifications are required. In fact, for the correct use of the AI-TRAFFIC-DEEP, AI-ROAD3D-DEEP and AI-INCIDENT-DEEP plugins, assuming it works with a maximum frame rate of 30 FPS with 1024 CUDA Core, you need: For a 1 CIF resolution:

  • RAM: 2GB (for the operating system) + 100 MB for each plugin; for instance, 8 plugins will require a total of 2,8 GB
  • CPU: Intel i7 or XEON, 1 core (or Virtual Core) each 1 plugins; for instance, 8 plugins will require 8 cores (or on a virtual server based on an XEON processor with 8 plugins, 8 virtual cores are required)
  • GPU: 6 GB of RAM for each plugin; for instance, 8 plugins will require a total of 48 GB

For a 4 CIF resolution, assuming that it works with a maximum of 4 FPS:

  • RAM: 2GB (for the operating system) + 200 MB for each plugin; for instance, 8 plugins will require a total of 3,6 GB
  • CPU: Intel i7 or XEON, 1 core (or Virtual Core) each 1 plugins; for instance, 8 plugins will require 8 cores (or on a virtual server based on an XEON processor with 8 plugins, 8 virtual cores are required)
  • GPU: 6 GB of RAM for each plugin; for instance, 8 plugins will require a total of 48 GB

For a 1080p resolution, assuming that it works with a maximum of 1 FPS:

  • RAM: 2GB (for the operating system) + 500 MB for each plugin; for instance, 8 plugins will require a total of 6 GB
  • CPU: Intel i7 or XEON, 1 core (or Virtual Core) each 1 plugins; for instance, 8 plugins will require 8 cores (or on a virtual server based on an XEON processor with 8 plugins, 8 virtual cores are required)
  • GPU: 6 GB of RAM for each plugin; for instance, 8 plugins will require a total of 48 GB

To make the precise calculation of the requirements necessary for the project, you can use also the spreadsheet at the following LINK.

The server is provided as a .tar.gz package that has to be installed on a Linux Ubuntu operating system following this installation guide.

Embedded Version - AI-Appliance

AI-Appliance is the plug and play embedded option, provided as an HW+SW integrated solution; the whole video analysis software is already pre-installed in the package. The AI-Appliance, allows management of both IP and analog (through DVR), simultaneous execution up to 6 video analysis plugins (with a 1CIF image resolution). Activation of plugins on the devices can occur dynamically through time schedulers or external input signals. Moreover, it is possible to activate outputs based on events (4 relays and 7 digital pins that can be activated through video event analysis and that can be configurated as inputs or outputs). To find out more, visit our website. If you want to find out more about the different versions,click here