So we have launched an open platform initiative; what does it mean and why does it matter?

The background

When we started Kinesense it was all based around the technology. Co-founder and CTO Mark Sugrue had just received his PhD in video analytics and he had some nifty ideas for video algorithms that meant video could be exploited with greater efficiency and effectiveness. Everyone else was focusing on real time detection – we asked ourselves ‘where does this technology work really well and solve a big problem?’ That’s when we decided to focus on video search or post event analysis.

Mark created an algorithm that was different to what was on the market, more robust to false alarms and yet would work even on poor quality video. Great idea, and perfect for police who have to sift through CCTV and video from shops and businesses to solve crime. It could be used to pinpoint when people entered a building or when red cars were captured on camera. The algorithm was perfect for analysing low resolution, low frame rate video, which was the main source of video that police had to analyse.

But that was the problem. All the video was coming from many different systems, in a variety of proprietary formats. So the first big challenge was enabling the video to analyse the video content of these different formats. Other issues to be addressed included how to enable anybody to easily create a search filter, how best to represent the video for intelligence purposes and how to ensure that the video maintained the chain of evidence for evidential preparation. We started with an algorithm and had to build a platform to make it useful.

Over the last eight years, Kinesense developed a video investigation platform based on the needs of police customers. The platform now has lots of great tools to exploit video intelligence and prepare evidential material. However, you can never stand still so it’s time for the next level of growth.

We know that there are lots of great researchers and companies who have built amazing algorithms for new applications and need a route to market. We know that video quality is improving and more can be achieved in video search.  Machine learning, and artificial intelligence developments are exploding supported by hardware developments. Applications which would make video investigations easier for police. So the next stage is about matching end user needs with technology.


How end user benefits?

Over the years, we have worked with international law enforcement community and helped them save enormous time sifting through video. Last year we saved 25,000 hours work for a single department (a potential cost of over €500,000).  However, as new sources of video data like social media and body-worn proliferate, new analytics are needed.

Not everyone needs the same thing. Different users will need different analytics to help them deal with video. Maybe your processing body worn footage which is non-static and you want to find faces and redact them. This will require a different analytic to someone with a static video looking for people wearing a red shirt.

The end user, either existing or new Kinesense customers, can select new video algorithms as options to the Kinesense platform. The basic video platform will be available to investigators but they can add new algorithms when the need them. In this way Kinesense will act as a shop for the latest video analytics developments.

As a technology, video analytics is at an exciting stage. The market is already aware of its benefits, but developments in artificial intelligence (AI) and deep learning mean that the technology can extend its usage to new applications and behaviors. We have already added third party face detection and recognition algorithms which are available to customers, and we have had interest from developers of audio algorithms to add gunshot detection and video developers to add lots of new analytics for things like gait analysis and crowd detection. These algorithms can make video investigations all the more effective.

How researchers benefit?

An issue developers face is that they don’t have a platform to test on. Also they don’t have a sufficient number of videos to test. When our Managing Director, Sarah Doyle partook in Joint Research Council initiatives like ERNCIP, a group analyzing best practice in video surveillance with a focus on video analytics. It was clear that on the academic side of things great video strides were being made in the field, but rarely were they being adopted. The researchers simply did not have enough video to test. There are data sets that exist but people often don’t know about them, and there are also issues getting end users to test and use analytics.

Kinesense are providing API’s to developers to enable them to develop algorithms and have them tested. We will provide the software platform for testing to academic researchers so that they can do this.

For any one interested in the work undertaken by ERNCIP, you can find a summary on Sarah’s LinkedIn page with links to valuable data

If you are interested in learning more, just let us know.

See coverage by Milipol magazine on this topic

Also see ASMAG article on this topic