unified-vsaas-analytics

3:12PM Dec 9, 2021

Speakers:

Sean Patton

amit kumar

Keywords:

cameras

cloud

analytics

gpu

hardware

carter

solution

integrator

problem

people

ai

processing

customers

run

dragon fruit

offer

big

support

capabilities

eagle eye

Alright, so is this Ryan? Is this not a panel? And we have some people joining you should be logging in. Alright, who do we have? Who was staying could? Or should we bring him back in? Yep. Em, Carter, and me. Gonna say hope this isn't a one man panel. I've heard there's a meet in Carter should be on in just a second Ian. There's Carter on this panel Good to see every buddy. No, we have to be in. Alright, and there we go. No, it's been a while hands. But uh,

glad you could. Yeah, schedule. Good to see you again.

All right, welcome everyone to our VSaaS and analytics, sort of unified panel. So really gonna focus on the unifying trends of you know, VSaaS, historically, yet VMS, SS and those types of systems have had more capabilities to offer, you know, integrated analytics and support through ONVIF. And, you know, their their own hardware on site. And that's where everything is going to be processed. But certainly, over, you know, the last five plus years, the cloud and its capabilities and processing and supporting analytics have kind of I don't say, even the playing field, but it's certainly something with a lot of VSaaS offerings, they're really starting to highlight the ability to to provide or offer analytics capabilities. And so we're going to look at this a little bit today. I want to give everybody just a quick, you know, maybe 15 seconds to introduce themselves. I guess we'll start with you, Carter.

And to see you. So background, I'm one of the cofounders of cameo. And we started when we met at Google saying Why can't you search video, the way you can search the web and 50 milliseconds. And a lot of our work is using analytics to make it overall more cost efficient. And tell the have the cameras tell you what you want to know.

I, this is a I met founder and CEO of dragon fruit. This is my third company I sold my first year I was second to NerdWallet, which meant I recently, the goal for us has been to build one product, one pane of glass, which has the most advanced BMS, and the most advanced AI analytics in one in one product. And that's that's what we have, obviously a chance to talk about that more during the battle.

And then Ian,

I'm in seamer I'm the head of product for open AI. And I'd kind of mirror that we're sentiments, we look at it as providing a single pane of glass where users can come in and interact not only with their video solution, but logically look at other events generated on site by access intrusion and any other data points that they can ingest, including athletics, and then

hands on scalar, Vice President Operations, Business Development and Eagle Eye. And I don't know what else to say, I can't stack up with these other guys. I guess they got way more better pedigree than I do. I'm just I'm just an Eagle Eye. That's it.

So I want to start off, you know, talking about edge versus cloud, because, you know, certainly we're seeing a lot of growth in smart cameras and AI analytics out in cameras, we've got shootouts and testing and performance is growing, you know, quite a bit. However, one of the one of the spots where there certainly are limitations is where you get to some of the more advanced analytics, whether it's behaviors, facial recognition, more, you know, higher or lower level details that can be offered, you know, maybe through cloud processing, we even see some manufacturers where, you know, it's a like a Meraki Verkada, where they're going to do, you know, people detection in the camera, and then they're going to analyze, you know, images in video to maybe a deeper level within the cloud. I'll start with you, when you look at, you know, the capabilities to what you can do in the cloud, versus maybe the ease and simplicity of doing some of the basic stuff at the edge. How do you guys balance that and really determine, you know, where's the best place for us to to process video?

Yeah. So take an example of our partners Hanwha. Right. So Honeywell has great AI cameras, but they work with us at scale on deploying dragonfruit as well. So why would you use them When, like dragon fruit when you actually have onboard, onboard processing, right, like aI chips right in there. And what we're seeing is that there is not only agent cloud, but there is also on prem analytics. Because even if you have two cameras, which let's say doing counting, very simple analytic, and they haven't field a field that is overlapping, no, on camera solution can do deduplication, because they're not talking to each other. Right? So one says 10 people or 10 people, there's actually only 20 people. But there's an overlap. So what we work with them, and I think this is going to be a trend is we have a base station, which is kind of this, which is actually a Apple platform, which talks to these hardware cameras, and of course, other bars and access cameras, as well does some processing on prem, and offloads most of the rest to the cloud. So if you compare with let's say, Verkada, or Meraki, where you do have, you have a limitation of how much processing you can squeeze in into a camera form factor, if you offload a little bit of it, and have these two on prem and cloud work together, you can actually go much further than a pure cloud or a Pure Edge solution. That's kind of what we think is going to be the way to go

in Yeah, no, I think that sort of leads into, you know, where unified V says, you know, maybe at least in the last couple years, maybe had an advantage where you have that single manufacturer, where you can offer that that hybrid approach. Carter, where do you see sort of as, as you're trying to get a non unified system versus, you know, if you're competing with a unified application? What are the challenges, you know, whether it comes from integration API, just the quality and what you expect from the metadata, where are the the challenges in not having a unified system necessarily?

Well, I mean, everybody wants to have a unified end to end experience. So that's always the goal. But a lot of times what we saw in the cameras were that they may have person detection, but then there's false negatives. So the problem with AI is that it's not a checkbox feature. You can't say, I do person detection, I do vehicle detection, I do counting, you always have to say I do those things. And I have 2% false negatives or 15%, false positives, whatever the ratio is. So the biggest challenge is making sure that the pipeline delivers the core value, which is tell you what you don't want to miss. And sometimes that means that you can't rely on kind of the weakest link in that chain. So you have to have ways to accommodate that. And we we do that, a variety of ways. If there's metadata available, it's always good to have metadata from where whoever can produce it. But unfortunately, that doesn't always supplant the need to, you know, look at the video stream, even even after you've gotten metadata from other sources. So I think the biggest ability to deliver a unified experience is making sure that you are open to all these industry standard cameras, all the industry, standard cloud services, whether it's Splunk, for logging or G Suite for your user administration, or any storage bucket that you want to use, all of these open interrupt things are key to giving kind of a great end to end experience. But you definitely have to make sure that you're not tripped up by the weakest link in that chain. Right.

I think

you can what's what's your follow up on it?

Yeah, I think that my humble suggestion is that instead of thinking about this as a multi system integration problem, I think we probably want to split the hardware and the software solutions separately, because I think that the hardware quality and, and an optics and so on has been has been going so much better than the software stack has been like if you think about the software stack from milestone, right. If I say to you, what's the best feature milestone added in the last three years? Right? Like, I'll give you guys a bit to think about our Genetec. Right, or open AI or Eagle Eye? Right? Like, you know, there isn't new innovation happening in the EMS industry. There is no new innovation happening in the analytics industry either. So I don't think the solution is like, Hey, how can we take these solutions that are kinda Okay, and stitch them together? I think what you really need is a revolutionary new system. I mean, from an IBM perspective, like we were all in fact, it interesting to be like Verkada, like, what the heck is that, but you can't deny the quality of the product that they have put out, right? I mean, there's issues around like, you know, lying and cheating, and like login, but the beauty of the experience is in having all of it integrated in one place, and we shouldn't deny

it I think that, you know, certainly makes sense in terms of, you know, as everybody is kind of talked about both in presentations today. And it's, it's in terms of what Carter said, you know, integrating those edge analytics is going to be important where you can, certainly understanding that that systems aren't going to be perfect. When you look at processing, and I'll probably go to you 100. Ian, when you look at processing in GPU in hands, you probably have a good sense in some regards, in that you guys run your, you know, your own database or data center hardware. Where are we at from a hardware standpoint, from processing, whether you need GPU and CPU? Certainly, you must have looked into, you know, offering some layer level of Cloud Analytics, in your data centers, you know, you do have some layers of Cloud Analytics, or analytics that are offered, when you look at the cost to do it, you know, maybe in one of your appliances, or doing, you know, some level in your data center, where, where are we right now, with the increased cost and GPU versus the capabilities may be out at the edge? Is there any shift that you're seeing, and that you would see to continue over the next couple years where the cloud becomes less important, the edge becomes more important, or vice versa?

That is a lot of questions. Yep. So let me just try to unpack a little bit of it. For those of you that didn't see my earlier bit, we do have an onset appliance that we call a bridge. So we do have some local processing, but there's no GPU in these devices. This is our older one here. But we do some basic what I like to call some basic analytics, we do some line crossing, some motion detection, loitering, some basic types of things. But where we're deployed, where we're planning to deploy GPUs, primarily is to continue to do so in the data center, it's much more efficient to do so there. So we can do the basic analytics locally on the bridge, see some motion CLI and cross send that clips and that event to the cloud and do some more advanced AI stuff there that requires the GPU. And that being said, we are exploring, expanding our bridge line to I know, we can talk to me about how many products we have on the price list already. But adding GPU enabled bridges, we're exploring that to see if it's worthwhile, we've got a couple of kind of Frankensteined up in our lab, and primarily there, it's because of latency. So a lot of the people have talked about, hey, you know, by the time you send an event to the cloud, and it goes back down, maybe there's too much time to do something. So we are exploring it but it's not on our roadmap, officially it just kind of stuff for us to to play around with here. But so the short answer is I see GPU in the cloud much more prevalent for us in the coming days. But we are playing around with localized GPU stuff, and it does it that's you know, you talked about supply chain. Right now GPUs are in high demand primarily because of cryptocurrency mining. A lot of those crypto miners use GPUs. So it's hard to get those. I was about to take one on my son's gaming PC the other day.

Yeah. And I guess that that would be sort of my yo, maybe a couple years ago, even, maybe GPU is more efficient. But you still think today, given the increased power capabilities on some of the edge devices, even looking at something like maybe a Jetson or even just, you know, some of the smarter AI based socks out there, do you think that GPU is still the more efficient way to go from a cost perspective in 2022?

Well, if you're gonna do it on the edge, certainly there are some issues mentioned Jetson those kinds of things. But since we're looking in the cloud, really the the GPU that we're GPUs, we're deploying optimized for, you know, hundreds of streams, or dozens, or depending on the throughput, that's really the, the direction we're headed. And honestly, I want about the edge of my technical knowledge there. I mean, we can get some other folks but that's what I'm being told. That's what we're doing.

Okay. And then in related to that, I mean, again, obviously, you guys, you know, look to push a lot of stuff out to the edge, whether it's, you know, through your cameras through your appliance, where do you guys sort of see, maybe even getting less out of the appliances more more really focused just on the cameras. And then also give you a chance to also respond to, you know, a mitt, you know, talking about a lack of sort of maybe development in features in that VMF space, you know, do you you know, tend to sort of agree and that's why you're looking at open AI you know, adding more analytics and some of the things you talked about in your session, you know, through your your sister company object video. You know, how the second piece but certainly from a processing and GPU standpoint, where do you guys see you know, the edge cameras versus appliance as being a better space?

So I've got a couple of takes on that. So One, I think, I think it's probably a mix, I mean, and so when we look at it, one of the one of the primary concepts that we use to drive our thinking and approach to it is what we call ubiquitous access. And so that really is that you have to have a little bit new partners where they are, but then also, they want ubiquitous access to that data. And so that's really kind of, I think, what users in the future are going to want. And so we think that the way to solve that is we have to bridge it across, you know, not only the edge, but also the server and then also to look at the cloud. And the reason is, is because it's going to be tough to come into a partner or an end user and say, yeah, we can give you Analytics, you just have to rip out every existing camera in your deployment. And so we really want to be able to come and again, support retrofit, which means that you have to have it either in the hardware or the cloud. But we also want to be able to provide more robust distributed deployments where they can have it across the cameras. And so a mix point, if you want to bridge cameras, though, you are going to need the cloud or the server to offer that are concerned a little bit with a cloud, long term is the cloud is the clouds a great spot. And it's good if a user's explicitly making an analytic rule, but when you talk about ubiquitous access, one of the things that we would like to see is that the stuff is just blatantly occurring in the background already, such that when you need it, if you didn't already have an explicit analog, or an explicit Analytics set up, you didn't lose it. And so, you know, to say, like, oh, well, you couldn't go find just personal vehicles when you search for him, because he didn't turn that on out of the gate. And we did that as a default setting, because we didn't want to provide GPU processing in the cloud for every one of your cameras, if you weren't going to do that. And so, you know, that's part of what we look at kind of as big as that ubiquitous access. And so it's likely to be delivered kind of across a spectrum of all of those blended together. And so to the, to the second point of no real new breakthroughs and innovations in the BMS, I, you know, to some regard, I do agree with that, which is, I'm not sure that there's, you know, the closest thing to maybe like a breakthrough search is a quote, unquote, Appearance Search. And maybe there's something else out there that we haven't figured out yet, but it's not, you know, you're not going to get kind of these like massive profound, I think shifts. And what that is about, again, the way that we respond to that is by providing natural, you know, ubiquitous access to what is there and then filtering it better, or sending it in a single pane of glass, making it more actionable, rising, the most important moments to the top, I mean, we're getting to a point where we're collecting so much data, that is not false positives, that it's still too much data to actually process and do something with. So you've got to filter that down further. Yeah,

I wouldn't apologize. I didn't want to purchase. Don't bring anything that one thing that I think we are missing, though, is like, your point is, like, look, what else is there to be built? Right? Like, you know, we have all of these things, we just make it more ubiquitous, more easy to access, there's nothing new, really, we can add, like all of the stuff out there, we just got to put it together? Well, I think that's true. Like, that's kind of my point I was making before about with Carter, which is like, it really isn't how do you take all these disparate things, bring them together? But how do you make it seamless? But you know, it's always cheaper, better, faster, right? Like those are the axes, right? So when we say dragonfruit, frontier is $1,000 per location per year, right, and you get hardware included in that. And you get 20 cameras included in that. And you get real time alerting, including in that, like, I mean, that is a fundamental connects improvement in cost, which is, you know, sitting on top of Apple.

Yeah, I think that's the problem.

Once I leave my thought, like when now that you have the end when chip, like all of this conversation about the supply chain go away, because now you don't have to bring your own GPU supply, you consider an apple supply chain and saying, look with the Amazon chip, you know, that you have in this box right now you can do advanced AI processing, at really low price points, because what did the Optimize advanced for video streaming? And machine learning? Guess what were the physical industry requires video streaming and machine learning. So I think if you don't take advantage of the new innovations that someone else out there is doing in consumer bring it to physical security and people that we serve, you know, then you're missing the boat, sort of you were saying,

Yeah, well, I think part of the problem is that advancing solutions that are cheaper, lower total cost of ownership, the answer is unlikely to be a whole bunch of new appliances, whether they're from Apple or other, that there's billions of dollars invested in Kubernetes as infrastructure for cloud native, even on premise. And what we think is a huge opportunity is to take billions from the big Tech Coast. make that accessible to all the on prem systems the same way you would have in the cloud. I don't

disagree, like we process 10,000 hours of footage in six minutes. You know, Kubernetes is the first step. And then you got to go build your own auto scaling system. Carter, I know your background very well. So Kubernetes is just like, that's like, that's like the starting point, right? So

you can run it on any? Well, it's just that you can run it on anything, you're not having to buy an appliance from one vendor. I think that's what the risk both on the supply chain, so you can

you can avoid it. And the reason you can't avoid it is because you don't have enough bandwidth. Like when we talk with customers that have like, you know, 10,000 16,000 stores, they don't have they have a T one line, right? Like, if you don't have any local processing, and you expect everything to go on the cloud, is that

right? That's what's good about on prem Kubernetes, is that you're not doing it in the cloud.

But then you want to roll out a $5,000 piece of code, right? Can you if you can do this,

this runs, this runs Kubernetes. It's not that basically you can run at any scale. So that's what is I think, a huge,

gigantic. Can you do 20? Camera object detection on that box?

No, but you can't do 100 On your box. So it's again, if you like it with a whole bunch of other devices.

Oh, no, this this this box does this box does 100 cameras at 10? ADP.

Yeah. Yeah. All right. Hey, hey, contact that dragon. Ai. Try it out. You posted and said you guys didn't work? And I will take it on LinkedIn. Yeah.

I think the other point is there is some practical limitation to hardware, you know, at some level, whether it's, it's 50 100. If you're talking about an airport that has 10,000 cameras, then you know, there's there is going to be some practical limitation for sure. That's, I think that's that's sort of worth the that is worth in something, I think we'll we'll end up sort of getting a follow up and kind of report and look into this of of the cloud and in the trade offs from both the pricing and hardware standpoint. This is the the debate of 2022. I think, you know, moving forward, you know, given the issues with with GPU and accessibility and bandwidth. Hans, do you have any input related to that?

Well, there's a lot of input I have, but the one thing I want to throw out there to the audience. And, you know, I love the debate. And you know, smiling, it's good, because I think you know that there's some truth that this industry needs more innovation. And I think that's something that all of us sitting on the panel believe and we all believe we bring it in in different ways. But the thing that I want to throw out there is that, you know, I don't think that there's any one tool, whether you're talking about Roger Stone video surveillance here, but it could be a hammer, and I tell my team, when you go into Home Depot, you look at the tool wall, there's 50 kinds of hammers on the wall, because there's that many kinds of nails in the world. And no one tool is the right solution for every situation. And so what might be right for dragon fruit, or cameo or Eagle Eye or, or open AI or genatech, or milestone or whoever may not be right for everybody. And so I think one of the things that we get into is I see the IPVM discussions on the site, is that somebody says, Well, I've got this great, you know, I got a little box to here, right? It's really good. Well, that wouldn't work at an airport. Well, of course, it wouldn't work in an airport, I wouldn't that designed to work in an airport, you know, and you probably wouldn't deploy Avigilon, you know, in the corner convenience store for two cameras, you know, so there's, there's just a lot of, do you need to think about your target audience as you're going through here. But the one thing that I will say that I think is, I think we all agree on the kind of the common thread is that whether you got a box from Apple, or you got a box around Kubernetes, or like us, we got a box that runs Linux, or whatever. I think one of the things that we're all kind of saying that I'd like to summarize and please challenge me if I misspeak is that we believe that there's a good mix of edge computing and cloud computing that are going to go forward. And by having some sort of a an appliance, you're not locked into the cameras. And you can leverage the cameras that your customer already has. A lot of times, maybe they have some capabilities icon. Why does work in Vanuatu? I think probably all of us do. You know where they can, you can get some AI stuff out of their camera, maybe you can do more in the cloud. Maybe you don't, maybe you don't need to. But I think that that flexibility is really the kind of thing that's going to help get this kind of widespread adoption, because the more flexible you can be, the more you can apply to the corner store and the airport, maybe have a different box. Maybe you don't do the airports but you don't do the corner stores. But I think flexibility is going to be the key going forward. So I'll end my rant there.

And how do you how do you think that Hahns is wrong?

Well, I can run out and grab my box too. And I'll show you I mean that's the thing is we all have we all are gonna have a slightly different take on the platform and the markets that we serve and we're all gonna you know, something we're all gonna be a little slightly biased in that way, but I think I'm more I agree with him more than I disagree with him, it's kind of hard to pick out because what because what I would say is, is the same thing, which is, uh, you know, there was a time when it was, it was pretty easy to go in and have a sale of, hey, look at these analog cameras you're running and check off this IP camera and how much better the resolution is. And then after that, you could go in and say, look at how much better the WD AR is on this new camera. And you had that kind of hardware and that software, really true feature set arms race. And now what we're all trying to solve for is that I mean, you can buy, you can buy a recorder system for 200 300 bucks from Costco or Best Buy it now. I mean, like, if you want a solution that just records video, there are plenty of things out there. And so when we're when any of us our sales teams are knocking on a customer's door, their buying decisions are more nuanced, they're exponentially more educated, their demands are significantly more sophisticated. And that's why, you know, to take it all kind of full circle into the heart of this conversation. It's why if if whether you're VSaaS, or whether you're, you know, local old school, DMS, if you aren't incorporating video analytics and other data points and forms of intelligence into your platform and taking a more kind of consultative cybersecurity and business intelligence role with the partners, you know, that those their pain points are that their prior system didn't give them the right alerts, or the management layer was such a pain in the butt that they didn't give their marketing teams access to remotely check video or like those are the problems that they're having and carrying with them. And that's what, that's what I think we're all I'm not hearing anyone on here. That's not saying like, that's the pain points we're seeing, and that we want to address. And so I think that's where we're all headed, we just all have a slightly different path to get there.

In terms of those pain points, Carter, I'll let you sort of address this one might end up running out of time anyways, but in addressing those pain points, do you find the challenge more with the the end user and who's buying it and understanding what the solutions are out there for them to even solve their pain points? Or is it more on that traditional security integrator who? Who is just locked into what they're doing? And they don't understand it? Where do you find more of the challenge? Maybe from a go to market standpoint, in education standpoint?

I even the big integrators are have digital transformation initiatives. And they know that people want more from their cameras than just recording video. I think they're evangelizing that alongside us. So we sell with integrators. And that's not I don't think the primary problem, I think the big problem is people seeing the entire budget for security broken out across VMSS, cameras, staff security staff investigations, you know, workplace safety, compliance and regulatory compliance. There's a whole big stack that a CFO understands. And I think, correlating the gains of knowing what's happening across your org to the big cost savings that are feasible if you do have that. Posture is what I think the big challenges is like selling to the CFO, you know, if you spent $10 million on your security systems, why can't you tell me? How many people were unauthorized entries yesterday in your West Campus? And that's like a embarrassing question that nobody wants to not be able to answer. But it's just been the fact because it's been too expensive to do things before. And this is the first time that it's become cost feasible to have kind of machines help you tell you the answers that you want to protect a facility. It same thing with a pandemic, it's put huge pressure, where people are used to their at home IOT automation, and they're like, What do you mean, I can't let the UPS delivery into my office when we're not there? Why can't that be an automated thing and those Well, you don't understand you have to integrate this access control and these video all they don't care that the expectation is higher now. So I think the combination of CFO looking at their total effectiveness of security, not just recording evidence, and then IoT pandemic demands have raised the bar and I think integrators are seeing that and selling that

especially when there's $100 You know, nest and ring and all those you know, their home systems can totally do that because they're they're one unified hardware solution so they're like why the hell can't my 100 times cost increase solution right my my home doorbell cam, right?

I think I think that's the key which is you need a solution that can be as simple to install as ring on the enterprise level, right? Like we have seen that before. Right? Like and in it. So you know, we have IPVM presentation as all of you all of us I do I have one day after tomorrow, and in 15 minutes we go and show things like Carter was saying which is Access Control integration, show things like liquid spill detection, which is the most advanced AI in the planet right now, but also show a video showing up with cameras being auto discovered. So we will show that for us. And I'm sure you guys will show yours as well. So within 15 minutes being able to go from like, I dropped in, you know, something in the network, and I am getting liquid spill alerts, right. Like that's kind of the magical transformation like ring dead like nested that we have to bring, and some of us are bringing into the physical security sector as well.

Yeah, really quick, I actually forgot this is actually a one hour panel, not our normal half an hour. So we got, we find that really, really quick a, you may need to do that demo in 11 minutes, or 10 minutes as those of our panelists who have presented with myself, I'm monitoring your panel, I'm going to set it, I'm going to give you maybe 10 minutes, I like to talk and ask questions. And if people want demos, I'm sure they can go to your website and contact you and get a much more view because quite honestly, what can you do in 15 minutes that you can't do in 10. And like it both are very difficult to from a demo show.

And Shawn? Yeah, now that we have more time, like I can say, so my I do a prep for the demo before I like that twice weekly webinars, like the links are over there. And the prep takes five minutes, because like you just run it through. So when talking, it takes a matter, I promise you, I you can hold me to the clock. And I'll show you all the things I added.

In so, so related to you know, there's been a lot of talk for this first half hour on open and supporting, you know, different manufacturer partners, devices and capabilities as camera socks that are out there. But specifically, as I believe most of you really only offer I guess, maybe open AI to some regards, would have it sort of a true quote unquote, end to end, your cameras, your analytics, your cloud, your, you know, whole platform. Why haven't like us, Carter, I'll just I'll go back to you, Carter, and then we'll sort of work our way around. Why not a cameo camera that has some level of intelligence and made metadata, you know, creating more stuff for you to look at, why not have a cameo smart camera that you can put out there not need a gateway, you know, have it performing some of those analytics control that whole process? Wouldn't it would seem that that would be a more, I guess, a efficient and trustworthy way to know what the results are going to end up looking like? Why not do that?

One, I just don't think we'd be the best in the world at that. Like there's so many good camera makers with all the different form factors and capabilities that our main promises like if you connect to cameo, you're going to get best of breed AI best of breed cloud best to breed cost profile, and we wouldn't be best at the camera. But apart from that, there's a huge amount of technical risk right now, that is, I think the top priority to mitigate for people that are buying systems because the AI landscape is changing so much over the next two to three years, with big shifts in capabilities. And so mitigating that technical risk doesn't mean like asking everybody to buy new chipsets and go up on a ladder and put a camera up. Because that is incurring a big business risk just for a rollout. And I think that's part of what our obligation is, is to mitigate that, and the companies that are doing that. I think they also have risks that are just like practical risks. Like if you take the Nashville bombing, you wouldn't have the video if it blew the camera off the wall. Unless the camera wasn't the place where you were recording your video. And, and so I think there's there's problems with a lot of dimensions to saying you're going to just do something only on a camera or only in the cloud. And that the biggest one, I think is just the overall risk profile of your project is radically reduced when you're using best of breed from industry standard camera makers that are already doing a great job.

Yeah, I guess I would, I would just sort of, I mean, partially joking, but also, you know, from some aspects of truth. So when you look at some of the startups that are coming into the space, and they're not all making their own cameras, obviously they're, they're, you know, using the JDM they're doing an ODM type of hardware solution. You know, what aspects that are maybe allowing them or or driving them to go in that direction. You know, versus you know, Cameo sort of has The backend stuff figured out, why not use quote unquote bolt on, you know, a, a cameo by Vivatech type of cameras in hardware solution.

Mainly, we're just following customers that they have good cameras in place, they just want them to tell them the right things. And so we're just in terms of our focus, we're just focused on the things that we can do uniquely well, and because that's if you don't focus on that, that's how we end up with a lot of mediocre vertically integrated solutions that because nobody's doing only what they do really well, and I don't see the demand there to come up with a cameo camera. You know, we it's just not I'm sure it's feasible, but it just doesn't seem like a smart business choice for us personally, just in pursuing what customers are asking for. They have a mix of access cameras, and Hunwick, cameras, and Bosch, and they've got all kinds of things that are performing well. And to put our adoption behind refresh cycle that involves getting up on a ladder is just not good for us. So that's, that's the main, that's the main driver for not doing that.

John, can I weigh in on that? A little bit? Yeah,

I was gonna go around the and have everybody sort of address that. So Hans, if you want to jump in on that next.

Sure. Cuz we do have cameras. I mean, we don't talk about him a ton, we do have some Eagle Eye cameras. But we support a bunch of others, just like, you know, most folks here do. And I'll tell you, it's, there's, you know, when you starting to deal with all the inventory, or one of our guys went to our factory slash warehouse, and they said, man, we got a lot of boxes for a cloud company. And you know, so there's, there's just the, the logistics of it is pain. But then there's also kind of like, this institutional inertia that customers seem to have or integrators seem to have, whether it's the end user, or it's the integrator about their cameras. And there's everyone, you know, I love access, I love Harnois, I hate this, I hate that. And it's almost sometimes becomes a religious war. And so to if you have, if you push your own camera too hard, then you lose out on all this other business that, you know, you're just not going to get. And I think to Carter's point of being able to focus on what you're really, really good at, is, is a very noble point. That's something I try to push us on. Yes, we get people that want to have the end to end Eagle Eye solution with our, our bridge, our switch, our camera, our cloud, that kind of stuff. And we have that. But I think the bigger push, right? No, the bigger push from just seeing what we sell, is that either they have existing cameras that are already installed there, you know, pick your favorite camera manufacturer, they don't want to go through and do it again, or the integrator has standardized on one or two sets of cameras. And they've tested them, they know how they work, they get the supplies, they got the mounts, they've got the warranty, they got the relationship, they got the discount, they've got all this stuff. And they're not switching off that camera for any reason. And so I think that for folks like us on the panel here, and I'll speak for myself specifically, that, yeah, we we have a camera and we sell it and we're glad to support it and you want to buy it, we'll be glad to sell it to you. But primarily, we view ourselves as more of the software, the cloud and now the analytics with our accuracy, recent acquisitions. But that's where we really want to focus. And then we'll use the cameras that are appropriate that our customers want to see.

Interesting. Yeah, yeah. It's very interesting, because we tend to not think about the hardware or the cloud equation at all. So we've tend to think of it as a networking problem. Right? Why What I mean by that is really, the question is what the choke point is, right? And the choke point is always the network, right? So you have your on prem, and let's say you have cloud, right? Like, previously, when there was no cloud, fine. Everything was behind the, you know, behind the firewall, so no one will talk about it. You have a local LAN, it's fast, it's fine. Right? But really, the question is the choke point and the access, because everyone is moving towards I want it on my browser and working from home, I need some sort of remote access and I VPN in, or do I have a cloud based interface? Either way, I would like to access my stuff, right, and stream it. And so there's three or four different models here, right? One is, I have a camera, I want to access to it. So camera to cloud, right? Now the problem there is really the firewall, right? Do you go out like Verkada from behind the network? Or are you going in from outside in to the network? Right? So that's like one scenario that you need to think about. The second you need to think about is like, Okay, I got a bunch of cameras, I don't have enough bandwidth. Right? Again, networking is a problem, right? Like, I don't have enough data that I can push up, right. Number three is I have a reason I have enough bandwidth, but I don't like what my camera feeds to go up like a casino. Right? Like of course I have a network to go out. You know, it's like a, like an optic of uplink but no video needs to go out. The reason I'm bringing these scenarios in is that this whole conversation around bridges and switches And, you know, you know, getting a platters actually goes all the way back to what's the environment? What's the networking environment URL? The reason that is a first point to think about is we're talking, this is a VSaaS. Panel, there's like, by definition, a cloud panel. Right? So you're gonna start with saying, assuming there's a cloud, this is a panel about cloud, like, what is the problem? The problem is that you have on prem stuff, and you have a cloud, and there's a bandwidth issue. And there's a hardware issue, because if you want to do analytics, you got to spend a lot of money to put GPU hardware here, or you got to put GPU hardware in the cloud. Right? Where are you putting it? Right? So So I think the question of the architectural design, like to take a step back really, is a question of not Do you run on the camera? Do you own your own cameras? Right, but assuming that all of us here in this call, are enterprise vendors, so you're selling to enterprises who always have deployments? The question is, if you have 1000 cameras, right? How do you traverse the network? So if a customer says I'm gonna open a firewall port, can you do camera to cloud some of us can, or some cloud to camera, right? Because you're opening up the firewall port? If you can't open that firewall port, can you do camera to cloud? Okay, so some of us can write, if you do have the requirement that the data not go up? Can you put a base station or some sort of sort of device? Great. By the way, I don't have enough bandwidth. Can you do a base station with on with computing? You know, yes or no? Right. So I think it's just the console saying we have to meet where customers are. But ironically, I think meeting the customers now is really more of a conversation about your networking stack. And under not everything is a factory networking environment, rather than your hardware environment, right?

potentially having your own hardware, your own sort of end to end solution. Make all of that easier.

It does not for enterprise deployments, because the opportunity that you have to go into a fortune 50 company and saying, Look, rip, rip and replace everything like all of us know that this is the problem the worker is having. They're having successes in the small business segment. But as soon as they're going up, they're hitting this wall, where people are like, awesome. Now, can you also integrate with my genetic system? And they're like, No, we can't. Right. So no amount of tech bros can solve that problem for you. Right. So if you are dealing with the, you know, neighborhood 711 Awesome. And I have full respect for that. Right? Like, you know, that is an important deployment. But you happen to have people on this panel that are all enterprise salespeople, right? Like they're all selling to large enterprises, who always will have existing deployments. And then the question is not so much like, Do you have a dragon fruit camera, but it is that given the camera infrastructure you have, can you do liquid spill detection? Can you do OSHA compliance, right, while showing me a real time feed of the video as well, right? Like that stuff we get for free trade, like VMS is free with dragon fruit. But but but the problem is that we are technically solving or not about which cameras you should take. But how do we deal with your networking situation?

That's fair. I mean, also, you're not thinking about hardware, but I think you've shown your hardware more than anybody on this panel. So you, that man has had more face time than in the other hardware?

As my as my my little face. But look, the the reason we do that is because we have this patent pending thing called split AI. Right? So we have this thing where

you talk about that in your presentation, I'm gonna, I'm gonna cut you off. You're nothing. You know, you're I agree, you're making your point. But that's, that's fine. Um, in terms of at Ian, you haven't had a chance to respond anything sort of different. You want to, to add or talk about in terms of the the advantages to having that unified system versus having to support you know, all the different analytics, you know, you mentioned you demoed, you know, Harnois and axes and supporting them. But realistically from from that dealer integrator standpoint, if they are in that space where they are looking for replacement, they are in a greenfield environment, clearly there, there would be a realistically there'd be some advantages to that unified offering, correct?

Yeah. Well, I mean, I think the cons made the, the joke about how many boxes they have on their shelves, even though they're a cloud company. I mean, what's the inventory that takes up the least space, you know, our VSaaS licenses, so we can stock millions of them in the warehouse. So yeah, we'd love to not have to have all these boxes in this hardware. But the reason that we do is because there's a bunch of partners that want that kind of the one throat to choke mentality, they do want an end to end solution. That is part of the kind of the promise of ease of use in a lot of ways is if you're using a platform that is all made by the same manufacturer and guaranteed to work together ostensibly you should be getting a better experience. And so we've looked at that pressure. But yeah, if we could just sell software licenses all day and not have to do hardware support and all that. But the reality is it, it also is easier for us to support and maintain, we would love to recommend our solution, not just because we make money on the box, but because it costs us less money in support in testing and engineering, to have a controlled environment that we know is, you know, guaranteed to work. And I think that that's you kind of call it out at the beginning of this. But that's exactly kind of the the one like really big benefit of Verkada, even though it is locked in is it is that Apple Magic of it's all the same thing, it's guaranteed to work together, it also saves them a bunch of development time by not having to care what anything else does, they control that stack, fully end to end, and that's great. And for the people that can buy into that all out of the gate, you know, I mean, I'm sure all of us can give a bunch of reasons why we think that you shouldn't do that, and to buy ours instead. But for some people that ease of use doesn't make perfect sense to them. And and they want to buy in into that. And you know, whether there's risks or rewards to theirs is what it is. But it's it does create a pretty elegant, seamless end to end solution if you're willing to stomach the other parts that come along along with that. And so yeah, we're gonna continue to invest in our own hardware, because we do at the end of the day think it does create probably a better and easier to support experience for our for our partners and end users. And I think that's why a lot of people get pushed pushed into that same space is to do that. But all of our partners have other preferred brands, you're taking over deployments that, you know, I think that's why I met said is you can go into these high end enterprise solutions and say, yeah, just rip out a couple million dollars in existing hardware you've deployed or you've standardized your IT, cybersecurity on Cisco switches, yank them all out, that's not going to work for us. And so you totally have to meet the users where they are. And I think that that's going to just become more common because everyone has stuff already deployed. There's very few true Greenfield opportunities nowadays. In fact, one of the areas that Verkada has seen a lot of success is specifically going after schools. And that's because so many of them have these grants and are building new buildings. And it's an easy place to get your foot in the door with regard to that. No,

that's, that's good. The last thing we'll probably get to, in certainly, if people have questions in chat, you can certainly throw those out there as well, as we, you know, look at okay, we have now camera partners, and we're going to work with best of breed and yada yada, yada all those things for having a more open platform. Where do we see it? I'll start with you, Carter. And we'll I guess the way I'm looking at my screen, we'll work our way around. Where do you see some of the, you know, plug in offerings. And as you look at like, and a cap at from access and their ability to run third party software now with deep learning, Chip capabilities. Some of the it was security and safety things, it's got a new name. Now starts with an A just slipped my mind. Where do you see that opportunity as a way of maybe offering a more unified type platform where you're still working with some partner cameras, but now they're running cameo analytics or something specific you're developing for, you know, facial recognition, or maybe some of the the the proximity alerting and stuff you guys offer? Where do you see that potential for, for expansion, the future of running your stuff in your partners, cameras?

I mean, I think it's good long term potential. The problem is, they have to see the market with deployment footprint that makes it attractive for us to target. So we've talked with them and are excited about the general ability to write apps for cameras, that's a good positive thing. But the things that aren't exciting is, you know, predicating our deployment on their deployment. And so I think once they get all the next generation of cameras upgraded, and they have the capability, we'll certainly target that. The the thing that I think is a problem is trying to replicate App Store models. I don't think people are excited about App Store models where there's a gatekeeper toll taker. And so that part, I think, is a problem. We ourselves are most excited about the cameras that are embracing, like a more open non SDK approach where you basically can run a container have access to the stream, and you can optionally use their SDK API surface but you don't have to. And I think that's super exciting. If if the camera makers start becoming more of like a general purpose compute platform, I think it'll blow open innovation because you're not having to retarget, you know, proprietary SDKs from each vendor,

is there a manufacturer like bid mate might be in the lead of that of that at this point.

There's some specialized ones from, like GPU chip makers, where they have like reference designs for this. And then there's a, there's progress on some of the Nazz. Companies that are also hosting containers and stuff. So I think it's coming, but I think it's ambiguous now, whether the camera makers will do that quickly or not. And that's, that's what I think we have to see a deployment footprint before it becomes something that you predicate your solution on.

Okay, yeah, Xena was actually that that was the former security and safety things are new name. So, Ahmed, in terms of supporting, you know, in putting dragon fruit into the cameras, you know, where do you see that in the next in the next year, two years? Where do you guys see that positioning?

In q1, we will be on a GAAP, we will be on Hanwa. And we will be on wash the surface Grid of Things, right platforms in beta for customers who need it for a full camera to cloud solution. So you take a hardware camera, which supports obviously, you know, they have different versions, some don't even have the app platform, you drop that in into a network, and boom, it's connected. And

then, then why you developing hardware solutions or software for the cameras illusions.

Because if this only works, if you don't want these cameras to talk to each other, because like I said, if you have two cameras that have a field of view, this doesn't work. So the customers who have one camera that they put outside the building, you know, and I can make you know, composite, and all they want this camera, we will have a solution for every customer who wants something like Verkada, right, with the Silicon Valley scale and simplicity in q1 For those who wanted. But for those who don't. We have as everyone has bored of hearing our apple based hardware solution.

Fair enough. And then Ian, obviously, you guys are in a little bit different space, being you know, you know, as much a hardware manufacturer is anyone else on this, on this next three days is going to be presenting but, you know, where do you you know, maybe see, you mentioned object video in, you know, your sister company in both your previous presentation, you may have mentioned it earlier in this one, where do you see maybe taking that object video stuff in and allowing having that software running, not just on your appliances, but maybe in the cameras, maybe even third party cameras that are offering, you know, object video by open AI, you know, plugins on their cameras, do you see that is is something that customers would would grab onto or, you know, it's just, it's pretty limited for what they're looking for.

It's, um, I think that a big part of it comes down to so I mean, you know, our core, our core soft, you know, we want to, we want to build all that into our core software stack the same way so that it can bridge across as many devices as possible. And I think that's where you'll see some of the breakthroughs is analytics that do bridge multiple cameras, as a pretty key area where there can be some benefit. So so we could take our server software today, if we just applied some some, you know, basic dev resources and make it run on a on a Honda or an access camera and those platforms that they're offering, and potentially bring along or analytics layer with that. And I think that I think that the the capabilities that that object video has, and the sophistication of some of those things is great. And then it could be out some, you know, many other analytics solutions for their performance, although, obviously any of us are going to say that about our solution. But I think the solutions ultimately are going to all be close enough in the long run, that that doesn't offer a significant competitive edge versus even just the embedded solution that comes on most cameras, and it probably is going to cost those up. And so it's just not going to become I think the dominant, the dominant market dynamic because of that, you know, there is still a cost to that at the end of the day, and the extra compute and all that's involved in, you know, pure cloud recording or still an edge edge or bridge or whatever device is probably the more cost effective and manageable long term solution that's still going to drive I think a lot a lot of that and it's just not going to make sense to to build that in into him or just won't be I think that the dominant technology, in my opinion, at least or at least, that's our stance today until the market shifts a little bit and we you know, we pivot next next year and

so on the next year's panel will call you out on

memory you said all we're doing is loading our software on other devices. It's

fair enough and then how Do you have? Once again, now we'll flip it around? How was Ian wrong in his statement?

Oh, he was so kind to me earlier. Thank you. And but you know, I don't know that he is wrong. I think that, you know, we've looked at this, we get people that like to play around with a Can we load our stuff on an access camera? You know, can we do those things, we actually have some open source software out there for camera manufacturers, called ie connect that camera manufacturers can install on their cameras and connect to one of our platforms, we have two different platforms, one that is can redirect to cloud and one that requires our bridge. And so we've looked at that, and the markets are just so different out there that I want to agree with the end of the day, if the market really pushes us there, and the cost, you know, all the economics get there, I think we'll, you know, we'll have camera direct to cloud and what analytic capabilities on there is gonna really be dependent on how much you want to spend on the camera. And so there's this interesting, you know, perfect storm of development, cost and timeline that we all deal with, right? You know, it's like, what you want, what you want to pay your timeline, pick any two, right, you can pick. And so trying to figure that out is what I think all of us on the panel are trying to do. But I see down the road, maybe not next year, but at some point, we'll see a lot more camera direct to cloud, especially as the cameras get more capable, they get more open. But I'll say the one thing that I haven't seen yet, that I think is going to be, you know, whoever develops this is going to really be ahead is the way to manage those things. The way to install those apps and your Carter said, nobody's really looking for the App Store is a gatekeeper in a toll taker, I think I hope I didn't butcher quote, too bad. But I do think one of the cool things about the App Store, no, for my iPhone, or Android or whatever. This is not my hardware. I mean, it's my personal hardware, it's on my company's hardware.

Like it should be ours. Yeah,

I wish. But the cool thing about the app store, regardless if it's Google Play, or iTunes Store, is it I can get updates to the apps really easily. If I want something, I can find it, I can install it, I don't have to go and visit some store. And you know, now granted, sometimes I have to pay sometimes it's free. But I think that that management of the apps for me on my phone, and then even some of the the enterprise stuff that you can use to manage apps for employees, you can push apps you can restrict, those are the kinds of things where I think an app store is really good. And I think that's something our industry needs to come up with maybe what are you guys doing after this? You want to bring something together?

Yeah, and I think that it's a it's sort of a good point in that the, the App Store model would be great. If if everybody could be in that Apple infrastructure, and the way it operates, unfortunately, right now will be seen with, you know, a Xena and with access a cap and with handle, it's all these separate stores? Yeah, and you know, nobody has really agreed on the one store and less. Well, really quick, from an integrator standpoint, my struggle with that it had was always well, it's not truly a store. It's just a rectory of different things. And you can't tell what works, there's no reviews, there's just a million things out there. And then if I want to buy it, you know, the value of hands is, you know, example of having that app store that does anything for him, but it's actually a store. And until somebody figures out how to build what is truly a store where I can go to my camera, hit the hit the you know, App Store button and pay for it in in turn it on right through it, which I think Carter's point was kind of like that's I mean, it's really you got to squint really hard to see that memory taking place in our in our industry for for a lot of different reasons and some legitimate reasons who's paying for it whose credit card is it going on? Who who manages any of the subscriptions going forward?

So what manufacturer gets to be the apple in that situation?

Right That brings us to their to the end of our time for now.