Cloud-based AI NVRs: Choosing the Best for Video Surveillance

‍

AI is playing an increasingly significant role in the video surveillance industry. AI-based video surveillance systems help customers reduce investigation time, set proactive alerts (e.g., identifying a person with a firearm), and overall save a substantial amount of time. With many vendors offering AI-based security solutions, it is crucial to understand the trade-offs between different architectures and how to select a solution that is:

• Cost-efficient
• Future-proof for new AI features
• Quick to set up
• High-performing in AI tasks

In this blog, we will explore each of these aspects and compare different architectures and some leading vendors.

How to Measure the Strengths of a Video Surveillance Architecture

In AI video surveillance, the ultimate performance of a camera is measured by the compute power (measured in FLOPS) per camera that the architecture offers. Since streaming video data 24/7 to the cloud is not feasible, all cloud-based solutions store and process data on the camera and push a minimal amount of data to the cloud to preserve bandwidth. This means the quality of AI performance is capped by the amount of compute power per camera that the architecture can offer. There are two broad categories of architecture:

1. Cloud-Based Video Recorder (CVR): This is a network video recorder (NVR) architecture with modern cloud features, making it easily accessible remotely and compatible with any IP camera. Coram Point is a cloud-based AI NVR that falls into this category.
2. Proprietary Closed Cameras: These are closed systems that process all data directly on the camera. In this architecture, the customer is locked into a single vendor and limited by the compute power on the camera. Verkada, Rhombus Systems, and Cisco Meraki are examples of this category.

Comparing Intelligence-per-camera

For AI video surveillance systems, the amount of compute (i.e., FLOPS) per camera is the most appropriate measurement of "intelligence per camera." More compute implies the ability to run larger neural networks and achieve higher accuracy results. Different AI NVRs can leverage different GPU accelerators. The CP-E-1 uses A2000 GPUs and is able to achieve more than 3.5 Tera FLOPS per camera. When compared to proprietary cameras such as Verkada or Rhombus Systems, which use chips from Ambarella (such as the CV25S), AI NVRs like Coram Point provide significantly more compute per camera. This means an end customer is able to get a much more intelligent camera system, and architecturally, they will be able to leverage future AI features without having to upgrade the hardware.

Strengths of AI NVRs Like Coram Point

• ‍Camera Agnostic: AI NVRs like Coram Point are camera agnostic, meaning customers can retain their existing cameras and add AI capabilities within minutes without having to replace any cameras.‍
• Higher AI Performance: AI NVRs are equipped with the latest Nvidia GPUs, providing significantly more compute power per camera compared to proprietary closed cameras. This results in more intelligence per camera.‍
• Future-Proof for AI Improvements: As AI capabilities continue to advance, AI NVRs are better positioned to support new AI features. Proprietary cameras become outdated quickly and cannot run newer AI features due to their older compute chips. This is akin to using a decade-old smartphone, whereas AI NVRs remain current with advancements in AI technology.‍
• Cost-Efficient: Since AI NVRs like Coram Point work with any IP camera, there is no need for customers to replace their existing cameras. This compatibility translates to a more cost-efficient solution. Furthermore, upgrading an AI NVR like Coram Point is very straightforward and takes less than 10 minutes, without any need to replace cameras.

In conclusion, understanding the strengths and limitations of different AI video surveillance architectures is essential for making an informed decision. AI NVRs offer significant advantages in terms of performance, future-proofing, and cost-efficiency, making them a compelling choice for modern video surveillance needs.