Looq AI Makes Photogrammetry Work

We talk with Lukas Fraser, VP of Product at Looq AI, about a camera-first platform that delivers survey-grade 3D models and automates utility workflows. We cover hardware design, accuracy claims, pole and cable analysis (for power lines), and why controlled capture makes photogrammetry competitive with LiDAR.

• Controlled handheld capture with four synchronized lenses and GNSS
• Calibrated hardware and encrypted storage with cloud processing
• Point clouds, panos, and ground-view outputs for engineers
• One-centimeter relative measurements for pole loading analysis
• Image-based detection projected to 3D for faster extraction
• Complementary roles of photogrammetry and LiDAR
• Subscription pricing with hardware, processing, and updates
• Integrations with existing survey and utility software
• Change detection to update models without resurveying

Transcript

Meet Lucas And The Product

Roopinder 0:00

Lucas, welcome to the show. Pleasure to have Lukas from Looq. That's probably that coincidence. Lukas from Looq?

unknown 0:08

No.

Roopinder 0:10

Okay. You're the I think you're the charge in charge of product, correct?

Lukas 0:14

That's right. I'm the VP of product here at it is pronounced Luke, but I do try to encourage Luke as it is closer to Lukas. Why not? Yeah, no, I've I've kind of been here since the beginning. I've known the founder Dom for a long time. And you know, we used to argue about, I was a big LiDAR guy early in my career, and we used to argue about LiDAR versus symmetry workflows. He thought you could get really accurate data sets with just the photos. I didn't think so. He he won out. He he proved me wrong. And so I had to had to join him in this venture.

Roopinder 0:45

Yeah, I want to talk about that a lot a lot more because that technology is fascinating to me, both photogrammetry and lidar. But where are you, Lukas? Where are you calling from? I'm calling in from New Brunswick. Okay. That's a long way from the company headquarters in San Diego, correct? Yes.

Lukas 1:02

So I I did used to live in San Diego for a while, and that's where I met the founder, but I'm from Canada and I'm I'm back here now.

Roopinder 1:09

New Brunswick. That's one of the maritime states, correct? Correct. By the way, so my history of photogrammetry goes way back. And I've always been a champion of it because it looks so good in theory, right? It just looks like it would work all the time. All you need is a camera and you can make these amazing models. But in practice, it tended to be a little bit more, how should I say, odorous, right? Not as accurate. And then when I saw LiDAR, you know, I was aware of LIDAR too at the same time. I thought, oh, LIDAR's got all the advantages. So accurate. It's it's it goes out, sees so far, it sees through things, right? This is like X-ray vision, right? This is what every everybody should have this. And how do you say that see that playing out in your company now? And are you are you using both technologies or are you favoring photogrammetry?

Photogrammetry Vs LiDAR Origins

Lukas 2:04

Yeah, that that's a great question. So um give you a quick overview on the product first. Fully, you know, first of all, we are a fully camera-based technology. So we have a proprietary hardware solution. And for anyone, you know, watching online, I know this is a podcast, so audio-based, but I'm holding up the hardware here. You can see we have uh four cameras in the front and then a built-in GNSS antenna, which is a very, you know, an important part of the process. You know, we often operate in exterior situations, so being able to pull in, you know, corrected GNSS along with the four cameras really is what helps us.

Roopinder 2:40

Lukas, hold it up again. When you light it up, it looks reminds me of Thor's hammer. You know, this is but this is the central product, right? Of your company. This is what powers everything. This is your unique selling point, right? It's also the I think it looks like it's also part of your theme of the emblem. Is that the am I looking at that correctly? That Luke's logo is actually that yes, it is. There it is.

Hardware Design And GNSS Edge

Lukas 3:05

Yep. That's right there, right there on the back. Okay. I think what actually sets us apart is the computer vision and photogrammetry on the back end, what people ended up really noticing us for is this hardware, just because it's it's kind of so different than than anything else out there. So it's it's often when people talk about look that the hardware is the first thing that they think of. But no, you are you are right, it is the foundation for everything because it if you don't have very you know reliable and high quality and you know the right amount of images, then that's where the photogrammetry starts to fall apart. So you you really have to be able to control the capture in a rigorous way. So if you're trying to pull in images from a variety of different sources, you're gonna get a variety of different results, right? So it was very important for us to create something that one, both you know, brought in the GPS and integrated it with the imagery, but also that we're controlling the quality and and the rate of capture, which is required for our for our stack. So a bit about our product suite, we have the camera, which is the the QCam, our kind of cloud-based processing technology, which is QAI, and then we have our web application for interacting with the data, which we call QApp. And the way a normal user would interact with it, it's they would power on the camera. It's it's very easy to use, intuitive. I mean, you know, most people are used to looking and taking pictures of things, right? So as you're capturing, you know, we have some guidelines, but in general, if you can see it, you know, in the imagery, then it's going to reconstruct in 3D, right? So you walk around, you're capturing the imagery, you upload it, it gets processed, and then we we make a variety of data sets. So point clouds, panos, we make like a ground-based kind of orthomosaic view, similar to what you would get from you know, drone aerial photogrammetry. We do some you know automations on the back end for for very specific workflows, kind of specifically surveyors and and uh you know make ready and polloading analysis. So those are our key segments right now.

Roopinder 5:14

Yeah. Hopefully, no, sorry. It sounds like you've totally swung over to the other side now. You're you're a you're a champion of photogrammetry and vision systems. Hopefully you didn't tell too many people to shoot you if you ever you sort of swung over to photogrammetry, right? Being a LiDAR champion initially, right? And you have a degree in in, I won't say degree of LIDAR, but a degree in geomatics, correct?

Capture To Cloud Workflow

Lukas 5:41

I do. Yeah. Yeah. I suppose you could consider me a bit of a a LiDAR trader now in a sense, but I I have a I have a degree in in geomatics engineering, which you know, we covered a lot of things, land survey, you know, remote sensing, ocean mapping, GIS, land administration. So, you know, a a kind of a well a well-rounded background. But I did spend most of my career in in remote sensing, reality capture, aerial mapping, starting with planes and helicopters, then to drones. LIDAR was always a big part of it, just for one, you know, like you said, the vegetation penetration is important. And that that is one thing that, you know, you can't do with photogrammetry at this stage. So there's there's always a place for both technologies. I kind of see this as very complementary in a lot of senses. But one of the main drawbacks to the photogrammetric methods was just that you wouldn't be able to get the level of detail on, say, on hard features, you know, curbs, retaining walls, anything that is often required, you know, for let's say like engineering purposes down the line, specifically like civil and infrastructure. Traditionally, if you've worked with, you know, drone photogrammetry, you'd see a lot of kind of rounding and smoothing of edges and features kind of joined together. And that's really, I mean, one of the, you know, the reasons that that I joined in one of our main selling points is that we're actually able to replicate that just being, you know, due to the quality of our imagery and and the refinement that we've done to our processing stack. So we're getting very sharp features on things like curves or tainting walls, and and very thin features like power lines as well, which are very important to our customers. So we've kind of taken photogrammetry as a concept, you know, like you said, you get a lot of detail, you get great colorization, and then we've been able to kind of improve the performance of these photogrammetric techniques for specifically kind of engineering and survey applications.

Roopinder 7:38

Yes. So your camera, is it is it only handheld, or is there also versions that can be mounted on, say, other platforms, maybe even drones?

Lukas 7:48

Right now we're very focused on the handheld capture portion of it because a lot of these projects that our customers are working on require a high level of detail. For example, if you're doing survey, you might have to get under trees and around behind things to be able to see everything. And if you're doing more of like an overhead utility pole-based workflow, you have to get around the back of the pole to see the tags. So it does lend itself well to the type of work that our customers are doing, but you can operate it at a very slow speed. So we have some people doing scooter captures and that kind of thing. But we do have plans to move into the mobile and more spaces in the future.

Roopinder 8:27

Got it. So that that device looks amazing. Uh uh, how many lenses and how many images is it getting? And uh then I want to talk about how it's able to process that and also determine objects from photos. So, first things first, how how many lenses and how what's a technology, hardware technology that that's using? Yeah, so four lenses, you know, four cameras in there that are all capturing simultaneously. We're getting around two images per second at the moment. You know, it's kind of an adaptive thing at the moment. I think if you're going kind of at a brisk walking pace, that's what you would get. They're all assembled by hand in our office in San Diego. So it is very different than let's say a consumer grade off-the-shelf camera. It's a you know, calibrated piece of field equipment. So we are, you know, focusing all the lenses individually. Then we're mounting it to like a robot arm, and we move the camera through, you know, like a fixed kind of trajectory in our our calibration room with the checkerboard targets. So every piece of equipment has a custom calibration associated with it that we can then monitor on the back end to sure it's not you know drifting away for where it should be, as well as push updates to it. So it's a very rigorously tested and calibrated piece of equipment. And and that's honestly what's required. You know, if you want to make uh high accuracy models, you have to have that kind of thing set in place.

Accuracy Claims And Validation

Lukas 9:56

Is that two frames per second per per lens per camera? Okay. And how many, what's the resolution? It's like an adjoint 20 megapixel camera, so combined from from the sensors. Okay. So you get a ton of images. How are they transmitted to where are they transmitted to? The cloud or are they are they a central processing unit nearby? So as you capture the the images are encrypted and stored on a device. So there's nothing that really could realistically be taken off of it. And then when you get back to the office or or wherever it is that you're working, somewhere with a you know, fast, reliable internet connection, you upload the data, goes to our cloud environment, and then within generally 12 to 24 hours, the the data is available to you. It's all fully processed. You log into our web application, and then you start interacting with it.

Roopinder 10:51

Okay. So I'm I'm sort of familiar with the initial attempts at re using reality capture for commercial projects. And of course, now they they stressed that you could use an ordinary digital camera. I think there was one example even of somebody just looking out the window of an airplane and able to get the entire Las Vegas strip from there, right? So I always always question that what is that it could be done? So you've taken a different approach. You emphasize the the hardware, it's commercial camera, unlike anything I've seen. Now, the trick here now is to convert those images into a 3D model. And that's where I always felt like the difficulty comes in because you have to register all those images, right? You have to create some kind of 3D model from those 2D images. That's a lot of work. I never thought it could be done easily. Just the steps that involved manually taking a picture, hundreds of pictures around an object. That looked like it was a ton of work, right, for a person. So you've that's all sounds like it's all automated now.

Lukas 12:08

Listen, I've done plenty of projects where we took images or LIDAR or other things and kind of were able to stitch them together through, let's say, great manual effort from very experienced technicians and a lot of you know ground control points. I think what we're trying to do here is to say, like, listen, let's give you something that is reliable and that performs in a very consistent way and reduces the amount of technical expertise required to get to that 3D model and frees up you know your technical experts to spend more time on the analysis or extraction or you know, whatever it is they do best, as opposed to trying to finagle a bunch of different data sources and get them into the format that you need.

Roopinder 12:54

All right. So I don't have to do that registration, it's it's just done. It recognizes points. There's I don't have to put any, is this way old school? I have to put things in the screen that help the registration, registration marks of some kind.

Utility Poles And Thin Features

Lukas 13:08

I don't that's no, no, it's it's all done, it's all done automatically. I mean, the one thing that you can do is you can upload field ground control measurements, you know, that a surveyor might get from a total station or or or like a higher accuracy GPS instrument, for example. But yeah, no, in terms of the the data registration, the 3D model generation, there's there's nothing that you need to do. You walk around, we're capturing data. We have some kind of recommended procedures depending on the type of job that you're doing, and you upload it and then you go on and there's your data right there. And if you want to supplement with other, you know, like I said, ground control measurements and whatnot, you can do that at a later date.

Roopinder 13:50

Okay. I see on your site that it claims uh very good accuracy. But I think you say one centimeter accuracy, which is quite something for photogrammetry models. Are you is that are you is that unique? Is that advantage or are you tops in that field?

Lukas 14:10

So I mean a couple of different things. So the the at let's say like the absolute accuracy, you know, the the georeferencing, where is this thing globally in the world is going to depend on your GPS quality. But let's say under under ideal conditions, you're getting a clear view of the sky, you're getting good referencing from your base station that we use for processing, then that can be, you know, kind of around the three to five centimeter range. So that's that's where it is. But intrinsically, you know, within itself, if you're taking you know measurements and and doing the analysis that an engineer generally would, that's when, yes, you're able to get kind of that one. So for example, uh a common application is a pole loading. So an engineer will capture a bunch of poles, and then they want to know all right, what's the height of attachment of all of the components on that pole? And how does this, what is this pole's position in relation to the next pole down the line for their tension analysis and and all the distance measurements that they are taking and extracting there have an accuracy of of around a centimeter.

Roopinder 15:12

I saw your news about the pole pole recognition software ability. And so most of the time, most of the time, when you see this application, application like this, you see very, I wouldn't say glamorous, because really just site surveys, but their entire surveys, site building sites or buildings in construction, not poles. So this is very specific poles and utilities and cables. So I take it you you Luke has quite a bit of AI with object recognition in this software. Is that correct? Because I gotta say, everybody can take photographs, but it takes a little bit of human intelligence or AI. Hopefully, AI has stepped in here to recognize all those things as real objects. Is that am I correct in saying this is all AI based and you've got a good machine learning, or you've you've taught it well?

Handheld Focus And Calibration

Lukas 16:11

So, I mean, we're very excited about it. I think you're the level of glamour maybe depends on the work that you're doing. But I mean, that's always the approach that that we've taken, right? We want to build things that are very useful for people in the specific workflows that they're doing. And that's really what the the Q poll module is, right? It's taken something that people do. They walk around, they capture information about poles, then they kind of manually click on all right, here's the base of the pole, here's the top of the pole, here are the two ends of the cross arms, here are where all the attachment points are, and then they reference that to their you know component libraries with all the details about the different types of insulators and and conductors and whatnot. So it's a lot of manual clicking. And the the traditional field capture techniques that they've used are kind of you know slow. So we're coming in and we're saying, all right, one, we're gonna speed up your fielding. You can now get this information really quickly, and then that's on the field side. And then the back office side, you know, you're having to click on the base and the top of a pole and the two ends of a crossarm. We have a ton of imagery, and it's very easy for us to identify that this is a pole and this is a cross arm in the imagery, and then pick the ends of that and project it into 3D, which is one of the advantages of being fully image-based. We have that relationship between the raw images because they're used to derive the 3D point cloud. So any objects that you detect in imagery can easily be projected out into 3D. It's a bit of an easier problem to solve than if you're working with LiDAR, for example, where you just have millions of points and you're trying to determine all the relationships based on geometry. The image segmentation is in general kind of easier to determine some of those things and then project it out into 3D. Now we're just trying to automate some of the things that you know slow people down and free up, you know, those expert resources and shift them more towards, you know, quality control and and the analysis portion, as opposed to kind of tedious clicking that has to be done, you know, day in, day out.

Roopinder 18:17

Yeah, it's quite a bit of quite a bit of work. This could help out with. So is is the end result of this application that do you throw away the photogrammetry then and end up with completely like CAD data, the CAD models, or do you keep both sets of data?

Data Security And Processing Time

Lukas 18:34

We keep both sets of information. So, you know, of course, it it depends on the user. I think sometimes users are just doing their extraction and then moving on. You know, you mentioned Autodesk. It's kind of a common survey workflow if you're making a topographic map, for example, to just reduce that point cloud down into a series of points and lines, extract, you bring all that information into Civil 3D, for example. What we're moving towards and what we're seeing the more of now that people have had this technology for a little while, they're starting to see the longer-term benefits of, you know, we have all of this information. And if there a change is made, you know, if somebody wants to attach something new to a poll, or if they, you know, redo a portion of the road, for example, we have all of the existing data and most of the extraction is done. So you don't have to start from scratch. We have both the you know high-resolution point cloud data sets and we have the models kind of saved in the back end. You can adjust a few things in the Looq platform, re-export, or you know, even kind of detect things automatically, you know, via change detection techniques. Here's what's changed about this environment. Let's update it, and then we can update your model and help you get back to you know the work that you do best.

Roopinder 19:55

How did you acquire the library of models to trip on? Did you is it just a big project to go around like I'm thinking of like a Google van with all its cameras and it's going around all the streets? Or do you do something similar to that with utilities, with with poles and and other fixtures?

Lukas 20:15

Yeah, it's a combination. You know, there's some existing libraries out there that we're able to pull on for a set of objects, but it only gets you so far. And then for the polls, for example, there aren't a ton of existing libraries out there for some of the more detailed components, different types of insulators and whatnot on a poll. So, but it's a combination. We had a good starting point with some of the tech libraries that are out there, and then we were able to build on that, working with our users on what they liked, what they didn't like, what they needed to add. And it's a work in progress. You know, we're still adding new stuff all the time, and we're going to be adding new stuff for the next couple of years. You know, there's lots to do as we expand into new verticals and and whatnot.

Roopinder 20:57

Yeah. Was there an existing market? Did you did you go up against software that the utilities themselves might be creating to do some of this? Okay.

Automation Over Manual Clicking

Lukas 21:07

We've definitely seen a large variation utility by utility in terms of their capabilities. Some of them do have, you know, very impressive libraries built out for detecting things. Uh, but I think a common theme is they they when it comes to the actual fielding and model generation, they often mandate, you know, for example, an end software. You have to get it into one of these three loading analysis softwares, and they sub out the actual fielding and that process to their, you know, engineering partners in general. So the engineering partners are often our customers, and we have integrations with the common loading analysis softwares. So we are able to help them kind of facilitate the transfer of information into the programs that they use already without having to go to them and force them to change their. Whole workflow, which, as I'm, I'm sure you're aware, going to a utility and trying to convince them to change off of what they've been using for years and years. Like from the fielding, you know, from the technology standpoint, it's hard enough, right? And then to go convince them to change into a different kind of final software is a whole other thing. So we're saying, hey, listen, you know, we can help you get more information faster and in a more affordable way, and it'll feed into what you're doing now. It won't disrupt, you know, your existing operations.

Roopinder 22:32

Uh-huh. I'm up in the Northern California area. We have PG and E, that's our utility. What is it in Southern California?

Lukas 22:41

San Diego Gas and Electric, and then there's SoCal, Edison.

Roopinder 22:44

Okay. Are they all customers?

Lukas 22:46

Or most of our customers are honestly their their service, their engineering service providers. So they they are aware of us and we do talk to them about things because they have standards. They want to make sure that the data that's being collected is done accurately and it supports their needs. So we've had discussions with them. But yeah, most of our customers are their service providers at this stage.

Roopinder 23:10

It seems like you've got a niche going. Might be a sounds like it could be a very big niche because there's a lot of utilities and a lot of utilities have a lot of structures. What's what's next? Or was that even the original idea? I think the loop was probably founded on more general surveying on photogrammetry, correct?

Lukas 23:31

Yeah. Yeah, we definitely didn't dive into it saying, like, all right, here we go. I mean, you know, our founder, he when he was doing his PhD, he was very focused on self-driving car technology. And then when we started the company, we saw an opening that we thought would make sense in the vertical, you know, construction side of things. But I had been working with, you know, with and around surveyors for a long time, you know, doing mapping. And the more we talked to customers, we kind of saw the need there for, you know, reliable and affordable and and quick capture techniques. So our first customers were pretty much all land surveyors. And then the request came in, can you guys reconstruct a poll? Like, how do you perform on power lines? Like those thin features are really hard to do. What can you automate there? And that's kind of where we how we ended up where we are today.

Roopinder 24:27

Was so it was it, you mentioned the self-driving self-driving vehicles, which I'm fascinated by. Have you taken a ride in any of the robotaxis, like the uh Waymo's that are out here in San Francisco?

Lukas 24:44

I actually have not.

Roopinder 24:46

Next time you're in the area, definitely do that.

Lukas 24:48

I'll have to give it a try. Yeah.

Roopinder 24:50

So were the cameras initially, like what were they going to be like driving along with vehicles and recording the entire scene?

Models, CAD, And Change Detection

Lukas 24:58

That that was just his initial area of research. The first prototypes all came out on cars or robots, or that's kind of what he was doing for his PhD thesis, was kind of computer vision in that space. But then it was just, you know, started the company through talking to customers. Say, you know, what are your actual needs here? You know, what problems do you have? What is worth you know paying for to you? That we ended up in the market that we are right now.

Roopinder 25:26

Do you find your company has to be sort of on the defense again? Again, I'm going back to LiDAR versus photogrammetry, but is that a tough argument to make? Could to convince people that hey, photogrammetry is is is usable and accurate? And is that do you do you come up against that a lot or not not? I I think once you show them models, they probably explain themselves, but do you get that chance?

Lukas 25:51

I mean, I encountered it a lot at first when I was doing UAV LiDAR, you know, you know, drone LiDAR back in in 2018. People were kind of initially thrown off by the overall concept. But I I think at the end of the day, it doesn't really matter to them. It's it's a quick discussion where they're like, you know, whoa, like based on what I know about photogrammetry, it's not gonna work. You know, take it, test it out, run it through its paces. I mean, what we know about the the survey and engineering community is it doesn't really matter. You can say whatever you want, but at the end of the day, they need to test it out, right? And they need to validate it against what is you know known benchmarks. And so that's kind of the approach we've taken. Like, you know, listen, don't take our word for it. Here's some data, play around with it, you know, take it out yourself, compare it to, you know, existing ground control and you know, whatever LiDAR captures that you have, and I think you'll find that it it stacks up extremely well and can get you what you need much quicker.

Object Libraries And Partnerships

Roopinder 26:53

I think your founders, at least the founders are probably aligned, if not friends, with Elon Musk, who always maintains that vision-based systems are sufficient, you know, that's all we have as human beings, and we we don't have radar or LIDAR, and yet we can make sense of the world. And he's insisted his vehicles have only vision-based systems. So I don't I don't know if he's uh pro uh funding companies like Luke, but he's probably lying with them, right? Because of vision-based systems, but so okay, so is my was my perception incorrect? It seems like you're bringing photogrammetry back, but seems like photogrammetry had actually subsided a little bit. Is that just my impression? Do you find that not to be true?

Market Fit And Integrations

Lukas 27:45

I think at one point, I mean, you have to go back a little a little bit further, probably even before my time in the industry, really. Uh photogrammetry was all there was, right? Right. You know, you were, you know, you were making kind of stitching together large, large models and and you were coming up with digital surface models, which were just the tops of the trees and and and whatnot, which was still groundbreaking at the time. And then light arcana came on the scene, right? And all of a sudden you're able to do, you know, from an aerial perspective, do large areas, even at a low density, you know, a couple of points per square meter, but still get down to the base of vegetation, which was uh exciting, I think, from a large when it came to more kind of granular, when you need really high density and resolution, you know, topographic maps the the existing technology prevailed, right? GPS and and total station. But drones came on to the scene, and all of a sudden, with LiDAR very specifically, you could do things like get the outline of a curb very accurately, but not as much from photogrammetry. You would need very, very high resolution images to get them from the air and spend a lot of time to get the resolution. So then, you know, LIDAR, I think when combined with imagery for mapping purposes, was very powerful. But LIDAR was kind of the standard in terms of the 3D accuracy. So I wouldn't say so much that it pushed photogrammetry to the side, but it was kind of the standard or the benchmark when you talked about, you know, getting accurate high-resolution 3D information. And that's what we're, you know, that's what we're coming here and saying, like, hey, listen, you can get that from photogrammetry. And as I said before, I still think that the technologies are complementary. And I don't think that this is going to come in and just completely replace LIDAR in every sense. I think there's still a lot a lot of great, you know, uses and applications for it. But yeah, I think we're saying, like, you know, maybe rethink what you know about what you can accomplish with photogrammetry. It can it can get the same level of accuracy.

Roopinder 29:59

To me, it always seemed as uh cost would be a big factor. I I know it is with Elon that he the reason the re the real reason he uses LiR is that cameras a lot cheaper than than LiDAR, right? Like they're in the tens of dollars versus LIDAR, which is in the now dropped down to hundreds of dollars, hundreds, but it used to be thousands of dollars. I know that that's also a big factor in the in his thinking. Is that also Luke's thinking too? That are you sell is Luke selling photogrammetry based uh photographry technology based on its lower cost? How much is that camera, by the way? That Torus Magic camera, how much is that?

Self‑Driving Roots To Survey Niche

Lukas 30:45

We don't need the LiDAR and we want to be just completely camera based that will help us fuel kind of our automation efforts on the back end. So that's that's the main logic. You know, if we thought that LiDAR would significantly improve the performance of it, you know, then then we'd add a LiDAR system in. So that's kind of that's kind of it. I mean, and then in terms of the pricing, we do we sell a subscription model, you know, all in you get camera automated processing, unlimited users in our web application for around $30,000 a year.

Roopinder 31:19

A year, okay. Oh, so you don't buy the camera outright. It's you it's you sort of rent it.

Lukas 31:25

So it's it's a bit of a change in logic, right? Because normally you you'd pay a large upfront amount for a piece of hardware, and then you'd kind of assign a a resource to process that data, right? And then you know, you would do extraction off of that. So we're saying, like, listen, we're handle everything. You get the camera. If something happens to it, we replace it. We have a very good return policy. So it is a piece of field, field equipment, you know. So we're kind of standing behind the the durability of it of it. And then you upload the data, it gets processed, you're not having to do anything there, and you just are are able to kind of dive into your analysis much quicker.

Roopinder 32:02

So that that fee that 30,000 a year, is that like a how can I say, am I limited to the number of projects or the amount of data that I can store?

Lukas 32:13

Or we do have a cap on it, but it in general, it's quite large. And the amount that you would have to do for kind of a typical, let's say, surveyor engineer is far exceeds what you know what you would ever accomplish. So it for all intents and purposes, it is an unlimited model for those types of projects.

Roopinder 32:33

It's like you haven't gone back. It's saying stop stop eating all the shrimp off our seafood.

Lukas 32:39

Okay, anybody I think we do enterprise deals for people that want to use it seven days a week, 10 hours a day. That kind of comes in more into the pool side of things, which of course is a bit of a different conversation. But in general, the usage is not that high for our engineering and survey customers.

Rethinking Photogrammetry’s Role

Roopinder 32:55

Oh, okay. Are you allowed to say, does a company say how many uh subscriptions it has, or how many cameras are at out there working?

Lukas 33:03

No, exactly, but I mean we've grown significantly in in the past couple of years. I don't have it off the top of my head, but we've had just pretty explosive growth in the past year and a half. So it's kind of changing month over time.

Roopinder 33:16

I'm really actually glad to see it happening because you know, like I said in the beginning, that the technology just seems photography just seems to be so immediately useful and accessible, right? That it there's gotta be some way, somebody's gotta make this work. Just like the process is cumbersome right now, you know. But I love the fact that I can just use a digital camera to get 3D images. I've been waiting for, I guess I've been waiting for a company like you to come along to make this useful. So yeah, good work. Keep it up.

Lukas 33:50

You'll have to get your hands on one.

Roopinder 33:53

I love to see it in action, honestly. That's that's amazing. And you know, people have to realize that you know, cost is commercial grade products are are worth it, that it's you know, you can't rely on you know, the the idea that you could just pull a little digital camera or something out of your pocket and use it for industrial industrial grade industry projects. I I never put much faith in that, but people have to realize they have to pay for good equipment. Engineers need good tools, there's no way around that, right? So tools then they're not cheap.

Lukas 34:30

Yeah, yeah. I mean, engineers deserve great tools because they do a lot of hard work and really nobody wants to waste time. Let's automate what is feasible and valuable to automate, and then let you know the engineers and and surveyors focus on what it is they do best.

Roopinder 34:48

All right, very good. Thank you very much. Lukas, thanks for making the time available. I really appreciate it. Very nice talking to you.