Video surveillance has long been understood in the security industry and many individual users have Rings or other home cameras for security purposes. But as of yet, there hasn’t been widespread innovation in video surveillance applications because it has been extremely difficult to bring together camera feeds at scale to run analytics. A recent webinar with Dominic Marino, Senior Solutions Architect at Storj and Cedric Verstraeten, Founder and Engineer at Kerberos.io, presented a very different, and very exciting future for video surveillance.
Kerberos and Storj came together to discuss how AI is changing video surveillance for corporate use cases. They discussed how the technology works and demonstrated how it can be applied in the webinar How AI-powered Video Surveillance is Transforming Security and Driving Business Growth. This article will cover the key points from the webinar, but a full demo on how Kerberos works can be seen in the on-demand webinar.
Today’s video-surveillance is limited
Marino shared an example of his own personal frustration when he attempted to monitor a construction project at his house with different cameras he had acquired over the years. “Each camera had a different app and I had to go in each separately to see the feeds.” Verstraeten shared the enterprise version of this pain. “Companies have legacy systems that are closed and very expensive if you want to extract data. And scalability and resilience is not there. You start with an NVR and add cameras until they are full, then you need another. But if an NVR box goes down, you don’t have access to your video.”
Today’s video surveillance systems are:
- Inefficient - Collecting too much data that is difficult to use.
- Closed - Systems are closed and isolated.
- Static - Systems can’t scale easily to bring together all the data.
However, Verstraeten said that “by changing to a cloud native system and adding artificial intelligence, these limitations can be eliminated.”
When video management is cloud native with AI it is:
- Efficient - Only collect data of interest, and streamline events.
- Open - Systems are open and camera agnostic.
- Dynamic -Thousands of systems can be connected and processed together.
These advancements can not only improve the cost structure and reduce the complexity of today’s video surveillance systems, but they can open up incredible use cases for applying artificial intelligence to video capture.
What can AI/ML do for video surveillance?
AI and ML in video surveillance stems from the YOLO (You Only Look Once) algorithm. YOLO is a real-time object detection algorithm that identifies specific objects in videos, live feeds, or images.
When applied to video surveillance, YOLO can enable capabilities like:
- Searching - Find specific objects or combinations of objects across unlimited video recordings or live.
- Alerting - Set alerts on specific objects, days/times, weather, or combinations of factors.
- Finding trends - Recognize trends at one location or across multiple locations.
According to Verstraeten, “applying and scaling AI is very hard for enterprises to do.” The algorithm continues to change and developing the expertise in-house to apply YOLO to a business use case, then scale it across the enterprise is nearly impossible with existing legacy video management systems.
That said, when done right, there are incredible use cases for enterprises and government organizations to gain value from intelligent video surveillance. Verstraeten shared examples of use cases in operation today.
Current corporate use cases for intelligent video surveillance:
- Investigation - Find all suspects with specific attributes and blur faces of everyone else.
- While this was an early use case, the ability to blur faces and maintain people’s privacy is an advancement being applied by law enforcement as well as by retail security teams.
- Safety - Alert when an employee isn’t wearing PPE or is too close to equipment.
- This is being applied at manufacturing facilities with dangerous equipment and at construction sites.
- Maintenance - Alert when remote equipment is damaged or not functioning properly.
- This has been applied by alternative energy companies monitoring turbines and even by billboard companies looking to monitor electronic billboards.
- Optimization - Learn trends for people traffic relative to day/time, store promos, etc.
- Retail stores from grocery chains to home goods have already begun applying this technology as well as hospitality venues looking to optimize flow. Even transit systems are using intelligent video to monitor bus usage throughout the day to optimize routes.
It is clear that many more use cases for applying intelligent video will be arising as this becomes more available and enterprises start experimenting.
How does AI infused video surveillance work?
Kerberos.io set out to solve the problem of scalability of AI models in video surveillance. They developed a system that is cloud native, cloud agnostic, and camera agnostic. Kerberos has an open source Agent, a Vault to hook into for proprietary AI models, and a SaaS platform called Hub that can handle everything in the cloud using Storj’s decentralized cloud storage.
Verstraeten went through a demo in the webinar showing how easy it is to set up and use Kerberos Hub. He showed how you can see all your cameras in one place to see their status and look at live feeds.
Verstraeten also demonstrated video search with deep learning and setting alerts for specific events. He gave the example, “A retail chain is running a promotion, they want to know how many people visit each store because of the promotion. They don’t need to view the video, the system can answer those questions for them.”
Verstraeten showed how powerful these capabilities were as well as pointing to features like heatmaps to show trends in flow. The best part was that Kerberos offered a free month of Hub for those that attended the webinar. Watch the on demand version to get access to the code and instructions on how to use it.
Why decentralized storage is a perfect fit for Kerberos
Verstraeten noted that having Storj as the backend storage provider is a “game changer for video surveillance because our video management platform is up and running at all times.” That means always having access to your feeds.
Marino explained how Storj makes that possible. “We have over 21,000 providers now on the network. Storj erasure codes the data across nodes and only needs the closest/fastest nodes to rebuild the file. And because we use parallelism, we could hit tens of gigabits of throughput.” And to explain the scalability of Storj, Marino made comparisons to AWS. “AWS can scale globally, but it gets incredibly expensive with regional replication. Storj is a 10x factor less cost than AWS yet it performs just as fast or faster globally.”
Why Storj was a perfect match for Kerberos to accelerate intelligent video
Marino and Vertraeten went on to talk about how Storj + Kerberos, as Verstraeten put it, “is a match made in heaven.” Both are open source, both are cloud native, and together they enable unlimited scale at the highest level of security. All while keeping costs low. And that combination makes it possible for organizations to innovate in incredible ways with video surveillance.
Intelligent video surveillance at scale is opening up incredible opportunities
The examples provided by Marino and Verstraeten made it very clear that the combination of Storj and Kerberos is enabling innovation at many companies and organizations that are already realizing value.
Many industries are already seeing benefits:
- Retail, Universities, and Government - Find criminals faster, keep citizens and employees safer, and reduce investigation time.
- Manufacturing, Construction, and Pharmaceuticals - Reduce workplace injuries, OSHA fees, healthcare and disability pay.
- Utilities, Distributors, and Government - Keep remote assets functioning, less field emergencies, and reduce maintenance costs.
- Retail, Hospitality, and Government - Improve store/business layouts or even city operations and transportation routes while improving efficiency and revenue.
With minds wide open the questions poured in for the speakers asking about system compatibility, how data leakage was handled, and excitement about getting started in their own environments.
To get the Q&A and the special offer from Kerberos for a free month of their enterprise subscription to the Kerberos Hub with unlimited cameras and 25GB/day upload, watch the webinar on demand here.