Thousands of cameras and the headache of being able to monitor them all at once is impossible without augmenting the intelligence of safety and security personnel. It requires a system that doesn’t sleep, gets tired or takes breaks. It doesn’t replace human security, rather, it takes over the mundane portions of security tasks and helps improve the accuracy of monitoring (making the human in the loop stronger) to improve response times in emergency situations. Providing operators with more immediate notification of issues can protect lives, minimize damage, injuries, and costs.
We have built our platform from the ground up to be accessible for security professionals. From retraining algorithms to fine-tuning accuracy, to deploying use-cases on streams, let us show you how simple machine vision AI can be.
Our platform comes loaded with intuitive tools that allow for fine-tuning algorithms to improve accuracy to training new ones for new behavior or situations. Break the chains of dependency on (software) updates from legacy solution providers and enable your business with powerful AI capabilities.
Scaling your security operation through AI is the future. With our platform it becomes easy to scale intelligence across oodles of camera-streams and locations. We employ the latest techniques and partner with industry leaders in the world of machine vision and artificial intelligence for the new age of safety and security operations.
A detector cell detects ‘objects’ needed for your use-case. The cell provides you with the object, its location in the frame, and the confidence level for that object. Training these cells on your data, repeatedly, will increase the accuracy and performance of your solution.
This pipeline example is built for Metropolai use-cases. By combining the below algorithm and application cells, we now have for every tracklet its position, speed and direction. Based on this data, rules can be defined to trigger events/alerts to catch unwanted behavior or events in real-time.
Measure the speed of individual vehicles as well as average speed of all traffic
Get alerted when a driver is driving against the flow of traffic while it is happening
Accurately count vehicles to start understanding traffic trends and possible bottlenecks in your road network
Classify vehicle types on your streets and roads such as trucks, cars, motorcycles, busses and more
Get real-time alerts of traffic jams or gridlocked streets and divert traffic for optimum traffic flows
Manage the response to dangerous incidents in real-time, ensuring swift safety and security measures
False positives are a major headache and waste resources. The technology of yesterday does not have the capability to learn from it’s mistakes. With Robovision AI, we save the recordings and use it to RETRAIN your AI to avoid it making the same mistake again.
By modeling normal behavior of which far more data is available we verify if observed behavior can be described by this model. In case it can not, it will be classified as abnormal behavior. In many cases there is limited availability of abnormal behavior compared to normal operations. Therefore it can be hard to model all possible scenarios and build something conclusive.
Pose detection uses an approach that detects key points (head, shoulders, joints …) of a human. These keypoints are correlated with each other to form a sticky like figure (simplified representation of a human figure). The behavior of a person can therefore be modeled as the way that this sticky figure moves and how its key points move/position with respect to each other.
While it could be possible that the detection of an object can be both normal and abnormal, it would be possible to differentiate between both scenarios by taking into account how the object is used. For an AI there is little difference between a crowbar and a walking stick, how that object is used is what matters.
Objects & Poses
Not only taking into account current pose and static patterns, but also how a person’s pose evolves over a certain time frame can describe their behavior and possible dangerous or illegal behavior
Planes, trains, streets or stores, we apply the above techniques to all. For us there is NO DIFFERENCE if we detect dangerous situations on an oil rig or in a bank, we work with our customers and partners to transfer their domain expertise to AI.
The Robovision AI store is the place to find new use-cases developed by us, our partners and even customers. It allows you to share your algorithms internally between your teams, departments and entities of your company. It can be configured to be both a private and a public environment.
In the cloud
On the edge
NVIDIA Metropolis makes sense of the flood of data created by trillions of sensors for frictionless retail, streamlined inventory management, traffic engineering in smart cities, optical inspection on factory floors, patient care in healthcare facilities, and more. With cutting-edge technology and an extensive developer ecosystem, businesses can create, deploy, and scale AI and IoT applications -from edge to cloud.
Network Optix is a software development company focused on creating cutting edge video management solutions. Network Optix flagship video surveillance product – Nx Witness VMS – is an open IP video management system(IPVMS) designed to discover, view, record, and manage IP video cameras and integrate quickly with 3rd party AI, analytics, I/O devices, and more.