Search for objects in video as you would in Google

Search for objects in video as you would in Google
As video surveillance footages increase, it becomes more and more difficult to find specific material in them. Imagine you had a ten-hour long video and wanted to find a specific car within that video. Would you have the patience to sit and watch the whole 10 hours of footage?

Or what if you could search for a specific behavior or an object in a video as you would search in Google? The Luxemborg-based Velocity provides the right solution for this. The company specializes in the creation of artificial intelligence (AI) solutions.
Speaking to asmag.com, Ivan Jacobs, Founder of the company said that the company has developed its core artificial engine Humano³, writing and implementing its custom proprietary deep learning algorithm for physical security.

“Humano³ is able to analyze video and video streams from security cameras. It is able to recognize and track objects, persons of interest, behavior, gender, emotions, movement dynamics, anomaly, cars, pedestrians, accidents, movement in relation to specific area,” Jacobs said. “Humano³ is able to simultaneously analyze multiple streams and track and follow persons or objects of interest.”

But what makes it special?

Velocity is able to provide end users the ability to create their own deep learning algorithms based on their own data. The customer can use them to analyze video and video streams together with the ones that the company provides out-of-the-box. Customers can continuously train the algorithms so that the system evolves and becomes more intelligent as they use it.

“The framework allows the management of the full life cycle of deep learning algorithm creation,” Jacobs said. “We have further automated all the manual tasks in deep learning development and decreased time to market from a few weeks to a day. We noticed early on that data is sensible and difficult to acquire.  Angle and lightning are always an important aspect that varies between projects. Using our solution the algorithm is trained on the same settings as it needs to perform on.”

Perhaps most significantly, the solution is open for integration and provides APIs. Accessing the API is possible in C++, Java, Python, Go, Ruby, C#, Node JS, PHP, Android Java and Objective C.

Behind the scenes of the solution

The overall architecture of Humano³ supports the analysis of bi-directional streams via public clouds, private cloud or standalone devices fully integrated pluggable authentication. The algorithms can be deployed on Android, iOS, raspberry pi, Linux, Mac and Windows.

“We leverage on the latest cloud technologies to distribute computation,” Jacobs said. “H³ performs lightning fast on CPU and GPU.”

He added that the company expects to stay ahead of the competition by innovating. It follows the developments in the evolution of AI and are early adopters and integrators of new trends. Of course, as the AI technology progresses, Velocity’s solution will also undergo systematic changes.  

“We will continue to innovate and implement deep algorithms even further in our solution. We are excited about the future of A.I. as it is our future,” Jacobs said. 
Share to:
Comments ( 0 )

asmag.com provides weekly and monthly e-Newsletters which include the latest security industry news, vertical solution case studies and product information.



Please key in code
Which Security 50 company will you place your bet on that grew the most this year?
Hikvision Digital Technology
Dahua Technology
Axis Communications
Avigilon