What about patent issues as a challenge? Is there a possibility on the concept may draw a patent challenge? Though I concede these days it seems like it would alomost be a given and maybe not worth mentioning anymore.
I think that [patent infringement] is the least of their worries.
It is the least of Shashi's worries perhaps. After all, as the founder, he is committed to the field and developing analytics, and would likely continue to try to grow his firm regardless of the IP issues that may be looming. With an attitude of "we will worry about that when we're successful", as you say.
Cuban is another matter altogether; he's only there for the successful outcome, and must decide whether the risk/reward ratio is sound. I can say from experience that one of the first questions out the mouth of any Silicon VC when looking at a startup is "What is patented and patentable?". Because it can mean a huge upside.
So, IMHO, Cuban had to ask about the technology used, and whether it could be protected, and whether it could be attacked. Anything less is borderline negligence...
Can't say how that conversation might have gone, but based on their general capabilities, they might soon have the need for a different conversation:
What kind of companies are in the risk zone for infringing your patents?
"If a company is generating metadata for use with analytics, whether it is for tripwire, people counter, 'object taken', 'left behind' or license plate recognition - yes they probably should have a conversation with us." Raul Fernandez
Though now it would be with a different Fernandez as well.
Cuban may be thinking of a synergy with another of his 'counting' analytics companies, MotionLoft.
Motionloft specializes in real-time vehicle and pedestrian counts for your locations - inside and outside. We provide data for shopping centers, malls, and commercial buildings; doorway counts, parking garages, urban street traffic and more!
Our proprietary, computer-vision enabled sensors, serviced by AT&T 3G, provide best in class service with minimal installation and setup requirements.
It is amazing that Cuban is so quick to do a deal again, considering that MotionLoft's founder went off the deep end and bankrupt the company, forcing Cuban to use his credit card just to keep the company afloat. The founder also defrauded various friends and acquaintances selling them non-existent shares in the company.
Netra appears to be more doing content based image search and retrieval in video than literal cross-camera tracking? It will be interesting to see what they are able to patent in this approach, but you can understand Mark Cuban's interest if they have solved the image search problem for fairly generic open searches with reliable results in real time.
There have been others doing this for narrow applications - one that I've seen first hand that works well is the OmniPerception (now Digital Barriers) Magellan video search engine. They built that to search, recognise and count advertising logos, brands and other image patterns in real time during broadcast video to help marketing folks quantify the benefits of advertising during sports events etc.
[Product note] Thanks also for the mention of Snap Surveillance John. Our approach is rather different. We apply computer vision techniques/ video analytics when first deployed on a large camera network to watch and automatically learn how the cameras relate to each other ... but then apply this in a UI for operators that helps THEM do the tracking of a subject very easily and reliably themselves. We didn't seek to build an analytic for fully automated cross-camera tracking (with the performance limitations / variability inherent in such an approach for large networks and busy sites) - but instead focused on a robust, 'one click', visual tool for the operator, using our learning to only show video streams in a view related to the current camera of interest.