What is NATIX?
As NATIX is gaining traction, we want you to truly understand how our work has evolved over time and what our core goals are in the near future.
In this crash course article of what we do at NATIX, we’ll start off by explaining why the current setup of real-time video analytics is problematic when it comes to cloud-dependencies. Then, we’ll discuss how NATIX can tackle such issues in an efficient and privacy-preserving manner to take cities and the private sector to a more conscious state.
The infrastructure is not an issue
Before we jump into the nitty-gritty details, take a moment to answer this question: have you recently experienced huge traffic what we now face in megacities such as London or LA? And, if yes, how often did you get the pleasure of a nice green wave? Unfortunately, the answer to this question is often negative. This problem has many different roots but at the very least why aren’t the cities trying to sense their streets at the micro-level and make faster and smarter decisions not just about their mobility but also other issues, such as public safety and other social challenges?
The problem is not the infrastructure.
NVIDIA estimates that by the end of 2020 we will have 1 billion city and commercial cameras which as just one type of sensor will produce approximately 1.6 exabytes of data per day.
The problem is neither the computer vision technology nor its maturity because today it is advanced enough to detect a wide range of applications.
It appears that the actual problem is the following:
Currently, if a smart city/industrial campus project wants to use cloud infrastructure for real-time video analytics, they need to send the data from the cameras to the cloud, process the data there, and then send the insight back to the place where the decision has to be implemented.
There are a couple of problems with such implementation. First of all, it is very expensive. Cloud services can cost anything between a few hundred to a few thousand euros per month per camera and, on top of that, you also have the communication infrastructure costs. Cloud is also quite complex to implement and the latency introduced is simply too long to support time-sensitive applications (e.g. autonomous driving). What’s worth mentioning is that such infrastructure endangers the privacy and security of the citizens’ sensitive data. Citizens/workers do not feel comfortable with having their faces and car number plates being shuffled around internet pipelines in these footages.
All of these aspects serve as a bottleneck for the wide-scale adoption of such solutions.
How NATIX comes into play
This cloud-dependency is exactly what NATIX is trying to minimize through what we name Opportunistic Edge Computer Vision.
NATIX Software Suite is a minimal software stack that can be running on any type of camera. It enables very easy deployment of any AI model across a whole network of cameras and turns the camera to an intelligent device that can perform on-camera video analytics. It also empowers cameras to exchange data, applications, or even computation power to perform functions such as data fusion independent of cloud infrastructure. This way, the cameras can collaborate and analyze a situation that individual cameras identify (swarm analytics). In turn, the camera network can deliver better detection, prediction, and planning.
In short, NATIX Software Suite includes:
- Event recognition AI deployment
- End-to-end security for the interactions that the devices are going to have
- Video anonymizer which is a module that anonymizes the data in the video in real-time before it is stored or passed to another entity
- Video optimizer which helps to filter the non-content holding part of the footage for optimization purposes
As a result, our solution is much cheaper, much faster, privacy-preserving and it enables the user to really utilize the entire infrastructure and the data that they have.
An interesting use case that bolds the advantages of NATIX is infrastructure-assisted driving in controlled environments. Here an exemplary customer could be a car manufacturer that would like to transport newly produced cars from a production facility to the parking lot. In the current setting, they are relying a lot on on-board car sensors but due to the sensitivity of this application, they also need the camera infrastructure in order to track and fix the vehicle trajectory or even cover the blind spots of the on-board sensors. Here, due to the time-sensitivity of this application, combining insights from multiple cameras must be performed at the edge of the network and communicated to the car locally.
Of course, NATIX has other use cases when it comes to city surveillance, mobility, or even safety at workplaces whose details we will share in the coming months.
Questions?
Your insights and questions matter. Feel free to drop us an email at hello@natix.io and our team will get back to you in no time.
DISCLAIMER: This post only reflects the author’s personal opinion, not any other organization’s. This is not official advice. The author is not responsible for any decisions that readers choose to make.
