These new AI platforms require expensive hardware just to review. I would like to see Anyvision invite the IPVM engineers to their facility for some hands on testing, unbiased of course. IPVM and Anyvision engineers can unite forces to produce a solid performance review.
Please put their product to a test - our experience with them was far from positive and we found out they claimed to have been connected to most of the cameras on site, while the pilot was on a couple of fixed cameras...
Do they offer a pilot system for customer's to try? I think so many integrators and customers have been burned by the promises of the performance of analytics (that are much simpler than facial recognition) that people will need to see it work in real time and not in a demo setting. Hard ROI numbers need to be proven for customers to purchase these types of systems.
We did an initial trial of their "Better Tomorrow" solution yesterday in the field for one of our police clients and they were very impressed with the facial recognition from a few of the street cameras. We tested with Avigilon 3mp fixed and 1MP PTZ. It pulls an rtsp stream that was very easy to setup. We are going to setup a "GPU box" onsite and run for a few weeks in the next month to continue to test the system but so far so good.
It's basically analyzing the streams comparing the white/black list of faces you've created or imported and then alerting based on that -- the alerts can act like an IO to a device or VMS or like a stoplight at a checkpoint boarder they have implemented (green light - ok to pass, red light - matched black list) They are working on VMS integration with the majors (think Genetec is done) but we have just seen their software so far.
Yesterday we basically fed my picture from linkedin and then pulled the rtsp stream from 2 of the Avigilon street cameras and then I walked with 2 other guys not in the system down the street and it identified me from that picture and we were able to download the clip of video into their software for replay -- the camera is mounted about 12' and it was id'ing my face 60' out looking down or to the side not up at the camera.
I think that will be the beauty of this system is the ability to id faces without the perfect image like we have been use to seeing in the past with facial recognition.
Paying projects pushed back our second round of tests but we now have a MSI laptop with Titan card analyzing 3 streams. Daytime shots are capturing more faces than night with just default settings but making some adjustments to increase the capture rate. I have a ptz that we have zoomed into an area and its really impressive how many faces it detects at a distance and angle that isn't what I would say is ideal. We have fed a list of faces and it's very easy add new faces directly from the camera detection. We have matched several people from day to day but don't have % numbers yet
Report updated with information from ISC West 2019, including a new mobile app and analytics appliances.
ISC West 2019 - Better Tomorrow Update, Mobile App, Analytics Appliances
At ISC West 2019 (IPVM Full Report) Anyvision demonstrated their Better Tomorrow 1.21 face recognition system. Using live video from multiple 2MP cameras within the booth, they demonstrated creating reference images with relatively low resolution captures:
They are displaying new small form factor appliances, targeting SMB and transportation, which they said would be available in 2019. Also, they have a Better Tomorrow mobile application that is currently supported on Snapdragon-based Android devices, which supports the full Anyvision face recoginition offering, and said face captures/detection events can be streamed to a central Anyvision system:
Access control seemed to have a much broader interest and TAM when I was previously at AnyVision. It seemed like it was mainly casinos, retail and law enforcement that cared about spotting a face in a crowd and were willing to pay the money AnyVision required to deploy the technology. I wonder if the first $70M of funding knows that AnyVisions field sales, partners and prospective customers have been telling them they want facial recognition for access control for about two years now. They've been way too busy trying to convince themselves and the market that they are AI company, yet they continue to only offer a facial recognition solution.
I'm very curious with how this news squares up with some key personnel continuing to exit the N. American side of the business. It appears Max Constant(Chief Strategy Officer - whatever that is)and Ben Ariel(Head of Commercial Operations) both recently left the organization. Taylor May; AnyVision's unofficial mascot, has removed all AnyVision affiliation with his podcast. Outside of the US, Lynette Pon and Jason Cook are also out.
Are these access control integrations true integrations now? Historically, they linked a face to a credential # and then sent that as a command to the access control system when that unique face was seen by Anyvision. Historically, this was not a bi-directional software to software database integration where users removed or added in one affected a change in the other. From the perspective of enterprise user management, this required redundant system management for card(face) holders. If this is still the case, this significantly limits the scale of the applications as very few organizations with a considerable amount of badge holders, who are linking to Active Directory already, will be interested in managing a second database manually now. Maybe IPVM can shine some light on the details here?
To the folks that just shoveled $43M into this company, you've had a combination of outstanding talent and technology and it still wasn't working. It has nothing to do with COVID-19. All roads point to leadership. They're strangling a golden goose. Stay focused on what the market is asking for and not what Eylon thinks is cool. He's not a visionary. Build deep, robust software integrations in the leading access control platforms - stop duct taping solutions together and calling them finished. The customers and partners aren't stupid.