The use of video analytics is experiencing a rapid surge, with the market projected to grow from $9 billion currently to a staggering $52.7 billion by 2033. This growth rate is significantly outpacing the overall video surveillance industry’s expansion, highlighting a critical opportunity for Video Management System (VMS) and Network Video Recorder (NVR) providers. To capitalize on this burgeoning market, these providers must find a way to offer video analytics directly to their customers or risk missing out on substantial growth opportunities
.
Until recently, VMS and NVR providers had three options for incorporating video analytics into their existing offerings:
Each of these options comes with its own set of challenges and in many cases results in a solution that is not commercially viable for VMS and NVR providers.
At first glance, cloud based video analytics seems like the perfect solution. No new hardware needs to be installed at the customer site and cloud resources are only provisioned to customers who pay for the capability.
Unfortunately, cloud based solutions bring their own set of challenges. Cameras must constantly stream video to the cloud which may cause issues for bandwidth constrained locations or sites with a large number of cameras.
The roundtrip latency to the cloud also means cloud based analytics solutions are not suitable for low-latency applications. High throughput access control and manufacturing use cases are just two examples of video analytics solutions that oftentimes run into issues when run from the cloud.
Finally, you run into standard cloud issues such as data privacy concerns and cloud resource costs. Some customer sites may not allow videos to leave the facility or prohibit the use of cloud resources for offsite storage. The cloud costs needed to support video analytics also add up when you take into account data transmission, storage and compute costs which usually require costly virtual machines that include GPU resources.
On-premises versus cloud is not a new debate and we have historically seen the pendulum swing from on-premise to the cloud. Recently we’re starting to see it swing back in the other direction as both solution providers and end users are realizing that some workloads are best run on-premises and at the edge.
The “standard” way to deploy video analytics on-premise has been to deploy x86 servers or appliances that contain Nvidia GPUs. Deploying GPUs at customer sites has its own set of challenges which need to be carefully considered by VMS providers and NVR manufacturers
Unfortunately simply adding a graphics card to an existing VMS gateway or NVR appliance is simply not possible. The VMS and NVR hardware is typically not designed to accommodate GPUs from both a mechanical and thermal design standpoint. GPUs require either a x8 or x16 PCIe slots and use anywhere from 75-350+ watts which generates quite a bit of heat that must be dealt with by the chassis.
GPUs are also expensive with the lower end models starting at $150. At this price point, it’s not feasible to add one to every VMS and NVR system which usually means you must deploy a separate dedicated server with GPUs for customers that want video analytics removing the opportunity to upsell existing users.
Finally, even if you solve the hardware issues you still need the expertise in-house to develop the video analytics solutions. The skillset to build video analytics applications differs from VMS and NVR software and computer vision engineers are in high demand.
If you’re a camera manufacturer in addition to a VMS or NVR provider then this option may not be as painful to you for obvious reasons. For everyone else, they want to maximize their existing investments in cameras that are working just fine.
Even if new AI enabled cameras are an option, oftentimes it is not the best one. AI enabled cameras carry a $100+ premium over non-AI models which adds up for multi-camera installations.
AI is also evolving at a much faster rate than improvements in other camera technologies so having the AI capabilities external to the cameras to support AI upgrades is an attractive option. The AI capabilities provided by cameras already lags behind other video analytics solutions, this gap is only going to increase over time.
At Deep Perception we believe that cost effective on-premise video analytics is the path forward for mass adoption and that VMS and NVR manufacturers already provide a critical aggregation point for camera feeds. With the recent introduction of low power, low cost and highly performant edge accelerators in m.2 form factors, VMS and NVR providers finally have the hardware needed to offer advanced AI solutions directly to their customers.
Deep Perception provides a full end to end video analytics platform that unlocks the full power of these new edge AI accelerators. Similar to how the m.2 form factor is optimized for a small hardware, our software stack is designed to consume minimal resources so that it can coexist with your VMS and NVR software stack. If you already have an events database and metadata search capabilities, we can easily integrate with your components otherwise we can white label ours to provide a seamless customer experience.
For VSaaS providers we also support AI appliances designed around a variety of edge AI accelerators that offer significant price/performance advantages over GPU based solutions. Our solution allows you to add video analytics to your existing camera to cloud and gateway based solutions by shifting the AI video processing to the customer premise while integrating with your cloud based VMS offerings.
As the video analytics market continues to rapidly expand, VMS and NVR providers must adapt to stay competitive and capture their fair share of customer spend. By leveraging Deep Perception’s advanced runtime that utilizes the latest edge-optimized accelerators, VMS and NVR manufacturers can now offer advanced video analytics capabilities directly to their customers resulting in higher margin software revenue. Contact us to learn how!