Video Security & Surveillance

Video surveillance technology has evolved rapidly. Analog CCTV and VCRs have been replaced by HD and 4k IP cameras and video management systems. Frame rates and dynamic range are increasing for better clarity in challenging conditions. Motion detection is used to avoid lengthy recording with no activity, and cloud-based services are eliminating onsite equipment while providing offsite storage accessible from anywhere.

Drowning in Data

Over 300 million IP security cameras are in use, with millions more added every month.  On average cameras generate 100 motion alerts and 2 hours of video every day, collectively that’s 450,000 Tb.  It is not surprising that security footage is used forensically rather than proactively.  Less than 1% of footage is watched, and even then, studies have shown that operator fatigue sets in after just 18 minutes.  The result is that motion triggered false alerts cost over $40 per camera per year in network, storage and related costs.  Perhaps more importantly, opportunities to proactively address threats are routinely missed.
Over 300 million IP security cameras are in use, with millions more added every month. On average cameras generate 100 motion alerts and 2 hours of video every day, collectively that’s 450,000 Tb. It is not surprising that security footage is used forensically rather than proactively. Less than 1% of footage is watched, and even then, studies have shown that operator fatigue sets in after just 18 minutes. The result is that motion triggered false alerts cost over $40 per camera per year in network, storage and related costs. Perhaps more importantly, opportunities to proactively address threats are routinely missed.

Video Analytics

Video analytics promise relief by identifying objects, and ultimately actions, of interest.  Unfortunately, video analytics for security applications have universally failed to deliver the combination of accuracy and value required to justify broad adoption.  The rapid development of computer vision based on deep learning algorithms is addressing the accuracy problem, but high equipment, network and cloud costs have prevented this from becoming a mainstream solution.
Video analytics promise relief by identifying objects, and ultimately actions, of interest. Unfortunately, video analytics for security applications have universally failed to deliver the combination of accuracy and value required to justify broad adoption. The rapid development of computer vision based on deep learning algorithms is addressing the accuracy problem, but high equipment, network and cloud costs have prevented this from becoming a mainstream solution.

Our Approach

Invision.ai is built on the belief that intelligence needs to be implemented where large volumes of data are created.  For security that means accurate, real-time, multi-class object detection and tracking running on camera.  After years of R&D we have developed deep learning algorithms so computationally efficient that they run on low-power low-cost generic embedded processors, such as the ARM A9.  The approach provides some compelling advantages over the alternatives:
Invision.ai is built on the belief that intelligence needs to be implemented where large volumes of data are created. For security that means accurate, real-time, multi-class object detection and tracking running on camera. After years of R&D we have developed deep learning algorithms so computationally efficient that they run on low-power low-cost generic embedded processors, such as the ARM A9. The approach provides some compelling advantages over the alternatives:

Hardware Agnostic

We work with industry leaders such as Ambrella and their ODMs to add intelligence to the highest value and most broadly used hardware. We can take advantage of hardware acceleration and can port to custom platforms where justified by unit volumes.

Upgrade Existing Cameras

Many organizations have made substantial investments in IP camera infrastructure that have years of useful life. We can retrofit these installations with firmware updates for newer cameras and low-cost processor upgrades for older cameras.

Network Efficient

Running object classification on camera reduces the volume of data to be sent to the cloud to a trickle: just metadata and high information content video. This enables large networks of cameras to be proactively monitored in real time.

Cloud/Edge Hybrid

Our system gets smarter over time as it gains experience. Edge cases are sent to the cloud, the algorithms refined and pushed to the camera improving both accuracy and efficiency.

Open Interfaces

We have robust, secure standards-based APIs for both camera and cloud services. Our technology is straightforward to integrate into your existing systems architecture and workflow.

Compelling Value

By avoiding expensive specialized hardware or big cloud server farms to delivery high accuracy in real-time. This enables us to be the value leader on a mission to make every camera smart.

Turn data into insight

Let us help you quickly eliminate false alarms while adding high value features like people recognition and dynamic heat maps.