April 01, 2021
How can we enrich and accelerate innovation with our partners and inspire them to take the next step in the evolution into more machine learning and AI type of deployments? This is as simple as enabling a more efficient way of building Video analytical applications, AI, and machine learning applications. The Video Processing Service (VPS) adds to existing MIP SDK capabilities and makes deployment flexible and easy.
Phoenix AI
Founded early 2018 by Laurent Renard, a former horse police officer with experience in building high-speed cameras for live sports broadcasting, Phoenix AI complements existing cameras with artificial intelligence and custom deep learning IVA (Intelligent Video Analytics). Instead of building their solution on cloud XY or servers, the startup brings everything back to a small electronic module that provides edge computing on the device. With only 10 watts of power consumption, Phoenix AI gives green technology to customers in the security, surveillance, Smart City, and live sports broadcast industry.
Increased business through collaboration
Phoenix AI first got in touch with Milestone for a major project with a well-known European railway company. The end-customer wanted to detect danger on railway crossings by identifying when crossings are closed and subjects are located in dangerous zones. They required the solution to complement their existing camera installation and be 100% deployed on the solution provider's infrastructure. "The end-customer wanted our solution to be 100% Milestone XProtect® compliant," explains Laurent, which is why we reached out to Milestone.
Bring innovations easier and faster to the market
The VPS provides a new way of integrating video analytics applications with XProtect. This is especially interesting for technology partners or solution providers who want to integrate their product with Milestone without changing the technology. Phoenix AI's architecture is Linux-based and not compatible with a Windows environment. The VPS acts as a shortcut for them to integrate with their existing architecture. Phoenix AI designed a plugin that runs on their platform and communicates with XProtect through the VPS to meet their customer's needs. Inserting their module into XProtect follows the same concept as inserting any video stream from an IP camera into XProtect. "All the customer has to do is bring our module out on the field and connect it to their camera network," explains Laurent.
On the development side, developers can easily understand the philosophy behind the VPS. Phoenix AI finds the framework well designed and easily understandable. Even though their architecture made it much less simple than expected, the integration took only one month and required only one development resource. Laurent explains that "for our solution, it's the best type of integration. It works out of the box."
Phoenix AI is an advanced partner from NVIDIA and is deeply in touch with NVIDIA's and Milestone's R&D teams. According to Laurent, when working with embedded artificial intelligence, you have to work with NVIDIA GPUs. The process involves combining the NVIDIA video pipeline with XProtect while bringing real-time metadata from the detection pipeline back into the Smart Client.
On the development side, developers can easily understand the philosophy behind the VPS. Phoenix AI finds the framework well designed and easily understandable. Even though their architecture made it much less simple than expected, the integration took only one month and required only one development resource. Laurent explains that "for our solution, it's the best type of integration. It works out of the box."
Phoenix AI is an advanced partner from NVIDIA and is deeply in touch with NVIDIA's and Milestone's R&D teams. According to Laurent, when working with embedded artificial intelligence, you have to work with NVIDIA GPUs. The process involves combining the NVIDIA video pipeline with XProtect while bringing real-time metadata from the detection pipeline back into the Smart Client.
"We supply a full solution"
SM_All by Phoenix AI adds artificial intelligence and custom deep learning IVA to existing cameras. Based on deep learning, their solution can detect various kinds of items and /or incidents, such as graffiti, traffic lights, cars, walking or running persons, etc. The customer sees live video streams inside the Smart Client with added analytics, such as bounding boxes on top of the video stream. Self-learning features enrich the solutions and make their customers' installations future-proof. Determined by the customer's requirements, SM_All learns by itself and starts detecting specific incidents already after 10-15 minutes. Phoenix AI can combine four detections in a row and monitor different kinds of incidents synchronically on their small product version.
Besides ensuring that railway crossings will become safer, Phoenix AI will support a great number of Belgium communities on a Smart Waste project. Many of these communities are already using Milestone, and Phoenix AI's integration of SM_All with XProtect will enable detecting the behavior of littering.
Phoenix AI is now also on Milestone Marketplace. Find them here.
Phoenix AI is now also on Milestone Marketplace. Find them here.
VPS stands for Video Processing Service. It is designed to easily fetch video feeds from XProtect and return correlated metadata feeds and/or a modified video stream back into the VMS system. Thereby, it allows to enrich video by adding analytical capabilities to devices that do not support these out of the box.
The VPS toolkit is an integral part of the MIP SDK and includes free samples for Windows and Linux operating systems, as well as samples using Intel OpenVino and NVIDIA DeepStream. The samples include full source code that demonstrate how to create a VPS. The VPS architecture and toolkit are based on the GStreamer open-source framework which creates pipelines of media handling objects. It is a well-proven, multi-platform technology that comes with a large community and a vast base of samples.
The VPS can run either on the edge on its own machine, on shared hardware, or even in the cloud.
The VPS was added to the MIP SDK in the 2020R1 release and is available for all versions of the XProtect VMS.
The VPS toolkit is an integral part of the MIP SDK and includes free samples for Windows and Linux operating systems, as well as samples using Intel OpenVino and NVIDIA DeepStream. The samples include full source code that demonstrate how to create a VPS. The VPS architecture and toolkit are based on the GStreamer open-source framework which creates pipelines of media handling objects. It is a well-proven, multi-platform technology that comes with a large community and a vast base of samples.
The VPS can run either on the edge on its own machine, on shared hardware, or even in the cloud.
The VPS was added to the MIP SDK in the 2020R1 release and is available for all versions of the XProtect VMS.
The VPS explained