Mixing Realities: A McCarthy, Autodesk, DAQRI Partnership

JORDAN MOFFETT: All right, so I know it's a small room, but we've got the mics on for the recordings. So is that too loud? Too loud? No. Good. Good. OK.

All right, so you guys are all here to talk about mixing realities with us. This is a, kind of, a try venture pilot between McCarthy, DAQRI, and Autodesk. So DAQRI on the hardware side, Autodesk on the software side, and us providing some of the industry knowledge. Jose, you want to control yourself?

JOSE NEGRETE: Hello, Jose Negrete. VDC specialist with McCarthy Building Companies primarily dealing with 3D modeling, rendering, animation, virtual mock-up, and AR RNT.

JORDAN MOFFETT: I'm Jordan Moffett, a VSC manager for our Southern California region. My role really is to make sure the team-- guys like Jose have the right tools, access to the tools, the time to do RND to make sure that we're continuing to progress in the industry. My background is I started my career as an architect in the industry and then transitioned into project management, and then got into VDC about four years ago.

So I, kind of, bring some of those practical applications to what we do and the aspect of having been out in the field or having been on the design side and just working through those pain points. So when we see that there's a different way to do business or a better way to do business, that's kind of what we try to focus on. I have a lot of ideas, but I don't really make anything happen. And that's where these guys come in to the equation.

PAUL CHEN: And my name is Paul Chen. I'm a director of product management for DAQRI. DAQRI builds wearable devices and software to help augment workforces. I've been with DAQRI two years tomorrow, which sounds like a short time. But in the AR wearable space, it's, kind of, a long time. We talk about years in DAQRI as dog years, and it's been a roller coaster over the last few years.

Prior to working at DAQRI, I was with an embedded software company for 13 years. We did operating systems for embedded devices. So DAQRI actually hired me to manage the operating systems of our devices. You may know DAQRI as the smart helmet company, and then a couple of years ago, we introduced the smart glasses.

I would've brought a pair, but the ones that I brought are all on the show floor right now in the expo. So if you go to the Vinci booth, V-I-N-C-I, they're showing our glasses with some of the applications they wrote. But now I manage some of the applications on our devices, and I happen to manage the one that McCarthy is using and we'll be talking about today.

JORDAN MOFFETT: And we'll have one more joining us shortly. Dave Tyner, he's the AR VR thought leadership manager for Autodesk. He's currently down in the expo hall. They have a pretty sweet multi-user environment setup down there that they're getting some press release around. So he'll be coming in hot and joining us upfront. So don't think it's some random guy when he does come in, all right.

So the agenda, so we're going to talk about the definition of augmented reality just to make sure we're all, kind of, on the same platform. There's a lot of terms and definitions that get thrown around, talk about the value of AR, a lot of the use cases that we see, and we'll talk about the specific case study that we endeavored upon with this-- during this partnership over the last several months. We'll talk about the current reality, some positives and negatives to what we're limited by with technology, future potential of you know where we think augmented reality can take us in the future, and then finally, just close it out with a call to action to all you guys and your partners.

So the definition for our purposes of augmented reality are simple. It's an image produced by a computer and then used together with a view of the real world. So you're truly augmenting your reality with digital content. The value that we see is so there's, kind of, four areas that we were looking at when we talk about the value of a AR, right.

So when you're thinking about-- you're going through schematic design early in the process using augmented reality to study different massing models or options for what buildings could look like, placement of buildings on a job site within the context of the existing reality. So for us, a ton of work we do is it's health. It's education. But there's hardly any Greenfield sites, especially in Southern California. So for us to be able to actually see what the buildings look like in context with the current campus and then show that to the owner and let them understand better what the design options are before we start going down that design development road.

JOSE NEGRETE: The other application we've tested is X-ray vision. So having the underground utilities that we've gone out and mapped at the site visible in the headset so that you can look at it in the field on site.

JORDAN MOFFETT: And then going into quality control, so as we progress into construction looking at what the future installations are going to look like in context with reality. So in the case that we're going to go through, we did this with an existing building, and some new equipment is going to go on the roof. It's all good.

But you can also do that on a new project where you start to look at the new-- the layers of construction that are going to come in after right. So you're constantly, kind of, looking at what does our coordinated model look like in context of what we've already built and then move it into the as built validation. So did the trades actually install per the coordinated model, or did something happen when we moved from coordination to fab drawings to fab and then out into the field. So for the case study, we really are looking at the two on the left there, the building massing and quality control.

So just to put it in context, one of things that we're trying to solve is to do these things in a better way. So we've always done massing studies and we've always done quality control and as built validation, but we've done it different ways. We've done it expensive ways like with laser scanning. We've done it with time consuming ways like going out and field measuring utilities as they're installed. So this is just to kind of put this in context of a way that we, kind of, do coordination traditionally, right, where we're marking up drawings and providing that feedback to somebody. But what if we can actually look at the markups in context with reality and be doing these markups in 3D real time.

JOSE NEGRETE: So moving on here, we'll see some of that computer interface. And as we go through these just keep in mind that McCarthy's initial R&D into AR involved much more complicated process with Unity 3D and basically creating a mini app every time we wanted to explore something on site. So luckily this works with 360 Docs.

So you'll see the interface here. You go to BIM Viewer. DAQRI BIM Viewer, it's a website. Log in with your Autodesk login. And it's basically authorizing DAQRI's app online to access your models, which you would have already uploaded with VIN 360 Docs.

So here you'll browse through your list of models, select the one that you're trying to explore, create a new scene, and then you'll see a familiar interface as you would with 360 Docs. And you'll see in the next video here this is sped up 400% or so. You're basically setting a 360 section box around the area you want to look at, add what's visible to a selection set, and then specify a location for the marker, for the landmarks that you will physically put on the site. Then this gets pushed to the headset, and you downloaded it-- you download it to the headset next.

JORDAN MOFFETT: And I will say I was remiss in really driving that point home that the platform that this is based on is BIM 360 Docs. So how many of you guys in the room use-- are on BIM 360 Docs now. OK, so the idea here is that you're hosting your models on BIM 360 Docs either through C4R and having them live on the Cloud or hosting them.

There is a model repository anyhow. So you basically have access to all your projects that you've already been invited to in Docs and then, of course, all the models that live within there. So it's critical that we don't-- we drive home the point that we're not introducing just another cloud. Because I know part of the frustration that we all deal with is having data in three, four, or five places.

So moving on from the computer side, this is the DAQRI headset interface. And for me, having been on the other side watching these presentations, it was just important that we put this interface stuff in here. Because we wanted to show that we've actually done this. We've gone through the steps. This isn't, kind of, marketing footage, but this is real life.

So here you have the model BIM. So this is, kind of, your first interface when you turn on the DAQRI, you have model BIM, which is where all the models are stored. And then you have a few options. So the first time you're actually loading a model, you scan the Scene Selection. So every single model that you saw at the end of that video has a QR code associated with it. So each of those QR's are unique to that actual model.

So you scan that QR code, and that actually starts the process of loading the model onto the headset, right. So now we're actually getting the model to the onboard computer. And then depending on model size, you have a certain amount of load time. And then finally, you get to the point where you scan the landmark.

So this landmark that you see is not unique to any model. So if you've got your landmark, regardless of what model you load, you place that landmark in reality exactly where you place it like Jose showed in that video virtually. So you're just matching the virtual width real world. So you have to have somewhere, somewhere on site, somewhere in your building that you know where that-- where that's located relative to the model position.

So a lot of this goes back to our guys actually going out and surveying the site and giving us-- there's little bit of prep work that goes into it. But you get a known xyz, and then you can locate your model based on that. And please also as we go through this just ask questions if you have them along the way. I don't want to-- we don't have the way to the end. Oh, see perfect timing.

AUDIENCE: [INAUDIBLE]

JORDAN MOFFETT: So the question was so, yeah, is the orientation known? Is the headset figured out? So basically that marker it's planar, right. And so depending on where you place it, so let's, like, the column in that example, depending on what side of the column you were to place that on it would rotate the model. So you would want to have it at the right height and also in the right plane.

AUDIENCE: [INAUDIBLE]

JORDAN MOFFETT: No, I don't even know if it would-- I don't know. Would it read it?

JOSE NEGRETE: It needs to be flat.

JORDAN MOFFETT: It needs to be flat. They're the expert.

AUDIENCE: [INAUDIBLE]

JORDAN MOFFETT: Right. Yep, and we'll talk a little bit about that because there was some playing with that location to make sure, hey, you load the model based on that marker. The model comes in and you can tell that it's slightly elevated. But at that point at least you can see, kind of, OK, it looks like we're about a foot up. So then you can drop it.

So I think that's probably some of the future development that we can get into is that the finite accuracy of that location. So for some of the things like this building massing study, that accuracy's not as critical as say that as built validation where we have some clients that require us to place objects, plus minus 1/4 inch of between as built model and reality. So the location becomes a lot more important in that sense. Good. All right, so-- [INAUDIBLE].

AUDIENCE: [INAUDIBLE]

PAUL CHEN: Yeah, I can take that. The DAQRI headset actually has a clear visor. For those of you who remember the smart helmet, we made it with a mirrored face. There was no reason to do that except the CEO then thought it was very cool looking and people did react to that. They said, wow, this is so futuristic looking.

When we actually gave it to people to use, we got feedback from co-workers saying if I'm working with this person who's wearing the helmet, I cannot see their eyes. And that really bothers me, because I don't know where he's looking. I don't have that eye contact.

So with the next generation hardware, we built the smart glasses. We purposely made the visor be very clear. So you can see the eyes of the person you're working with. Now you bring up a good point. When you go outside, it's very bright.

We are shooting virtual content onto those displays. It's like a computer screen, and when you take your Kindle or even your computer outside it's very hard to see. So what we've done is we've built a little clip on basically sunglasses over the visor, and it works quite well. We were on site with its outside looking at the hospital. And the virtual content looks very bright against the background.

JORDAN MOFFETT: That question may have been founded off the HoloLens, so not to bring up bad words around our DAQRI partners. But that's one of the issues, right, is, well, you've just cut out half of our environment. And even then even inside the building as we go through construction, you may not be closed in so you're always going to have potentially direct light. So the rooftop study, this outdoor study, I think both came through very well.

So let the video roll. So again, just wanted to show kind of where we're at in the RND and give everybody a flavor of the type of content that you can bring back. So ultimately, we want to get to the point where we got see what I see type of technology showing people real time. But in this case at least we can go out, grab a case, grab some examples of maybe cycling through three or four different models to be able to look at the different options, whether we have a two or three story building that's, kind of, more wide taking a more space. Yeah, the scale gets a little bit-- a little bit crazy.

But anyway, so the idea here is that you can just you can capture these videos when you're out on site and pull them back, bring them up with the larger design team, the owner, and start to say how do we want this facility to actually look. And I think the other goal is for us and the frustration lies in the owners obviously that we work with don't do what we do every day. And so to fool yourself into thinking that they really understand what the projects going to look like by reviewing the plans, elevations, sections is just that, right. We're just fooling ourselves.

So I think if we can actually build a building that the owner goes out and when they see the complete product they say that's exactly what you showed me two, three, four years ago when we did those site studies, I think that's a huge win for all of us in the industry. The second part of the case study was the roof rack example. So there's the new building that you just saw. And then we have building three on campus, which is an existing building, basically has some retrofits to the mechanical, the cooling system.

So we have new roof racks and then some of the mechanical piping that's going to be on those roof racks. So one of the cool things about this case study was the architect got the as built documents. They as built the model to then be able to go and design the retrofit. So when they as built the model, they followed the as built drawings that were x number of decades old. And there was actually a parapet wall on the roof that was about a foot off in the as built.

So when we went out there with the hologram and we were studying the roof rack in context with reality, we saw that the roof rack was actually into that parapet wall. So we were able to actually take that back. This is a trade partner model. So the mechanical-- the mechanical trade partner was actually able to go change their model, re-validate through Navisworks' coordination that we weren't clashing with anything else that we thought was out there and then re-validate through augmented reality that the proposed solution actually works.

So pretty cool to get a win, an early win on a project like this. And so you can, kind of, see where that value could come into play as you're building a new building and, kind of, always just looking out into the future of, hey, if we go pre-fabricate this item is it actually going to fit when we bring it into building. So now we want to, kind of, transition to what's the current reality, where are we out with the hardware, software and just in our work environments, and then we'll talk about some of the future state.

PAUL CHEN: With the current version of our smart glasses, we made some very conscious design decisions. One is on the input methods. So you put on the smart glasses, we've been doing demos for years now. We have a competitor, and they like to use gestures.

So the first thing that people do when they put on our smart glasses is in order to do something, they start doing this. We have this design philosophy that we're building our device for people who are doing work. They're doing tasks. They're typically using their hands.

They may be wearing gloves. They might have grease and dirt all over them. They've got tools in their hands. They're twisting knobs, working on machines, carrying things, laying down materials. Their arms are very busy.

And what we didn't want to do was give them the added burden of having to do this a lot of times. Think about how many times you click a mouse in the course of your daily work. Imagine doing this for every time of those and actually you sometimes have to click two or three times to get it to take. I've done maybe a dozen clicks standing here, and my arms getting tired. So we didn't want to add that burden to our workers.

JORDAN MOFFETT: And everybody stares at you when you're doing it too, and that's the other thing.

PAUL CHEN: They're already staring, because you're wearing this [INAUDIBLE]. We live in the 21st century where people are used to it. So we made a design decision to make our user interface completely hands free.

So we have what we call a little reticule. It's a white dot, and when you look through our smart glasses the white dot follows your head. So you control the dot by moving around. When there's a user interface element like a button or a menu, you activate by looking at it and then what we call dwelling on it for less than a second. And then you've pushed the button or selected the menu item.

People pick this up very quickly. Even down in the display booth, they started doing this. And I say we don't do gestures, and they go, oh, there's a white dot. And within 15 seconds, they're navigating our user interface. I say click on that. Oh, yeah, I did that already. So that was one bonus. People could still do work, climb ladders even while wearing our smart glasses. Question?

AUDIENCE: [INAUDIBLE]

PAUL CHEN: So the question is is it eye tracking, and no, it's not. One of the reasons is eye tracking cameras are added expense to the hardware. Also with different eye positions of each user, it's very difficult to get a good eye tracking position. So this is totally based on where your head is pointing. So you control it with your head motion.

Another good feature that people gave us feedback about was the ability to work offline. Now of course, your models are stored in BIM 360 Docs. They're in the cloud. That remains your authoritative source for the models. Our hardware doesn't touch the model at all. We don't change the model. It stays there.

But of course, if we're going to display it for you, we do have to access it. So as Jordan mentioned, you download the model to the device. If you know you're going to a job site where there is limited or no Wi-Fi capability, you can then store the model locally to the device. In the user interface, we have a set of buttons, one of which is download and save the model.

Then when you're on site, when you launch the app, Jordan showed you how you can scan a QR code to download a model. The second option was load a saved model, and that way you can just refresh one from the stored memory question.

AUDIENCE: So how many models can you [INAUDIBLE]?

JORDAN MOFFETT: So the question is how many models can you hold? And I would counter that by asking how long is a piece of string? And it depends on how large your models are. There is 64 gig of hard drive on our device-- about a third of which is our operating system and the applications and about another third of which is the recovery partition, which in case the first one fails, you can back up. So there's about 22, 23 gig of hard drive space.

Now that's one limitation. Going to next question, can we actually render a 23 gig model on our glasses. The answer's no. Think about your designers. They're using these humongous Alienware or other desktop machines with 14 and video processors in them.

We have an embedded processor in our device. It's a laptop processor. It's an Intel Core M7, but it's going to chug on a large model. So as Jose mentioned, you want to segment down what portion of the model you want to view. That's not only to optimize the download speed of that model but also once we have it on the device that little Intel processor's chugging to get it out onto the glasses.

AUDIENCE: And how big was the whole building or polys we were talking, right?

JOSE NEGRETE: Yeah, it was like 500,000 polys still. And you saw it there's a sensor in the interface, so when you're defining a selection of geometry that you want to upload, there's a little HUD thing that shows you if you're well beyond the reasonable poly count or if you're within it. And so we're actually within it, but it let you go higher than that. You'll just experience a lot more jumping and jittering when you try to do the model in the headset.

So this again current state. We're looking at anchoring. So if some of you noticed in the video for the building facade the 3D surface where the marker was in 3D space was actually several feet in front of the actual wall where we set up the marker. So anchoring is something that's important for now, because you have to go out on site and have somebody set a control point for you to anchor the 3D model.

And in this case, we had an as built model that was actually in a different location than what our 3D content had. So we had to shift the model on site. Other things to take into account again is the model complexity. So like we were just talking about, you have to section of your model into blocks or like using the grids of your model to explore more than one little area of your project.

JORDAN MOFFETT: But the nice thing is that if you do that right, you can have those multiple models loaded with the capacities so. And then using that same landmark, you can just move the landmark from location to location. So a lot of that's what we've been, kind of, playing with and trying to push the envelope as much as possible.

PAUL CHEN: Sensor and other hardware limitations are also part of our current reality. The smart glasses that we have today were a direct evolution of our smart helmets. And in order to get the smart glasses to market as quickly as possible, we basically use the same exact hardware configuration that was in the helmet, same cameras, same IMUs, same processors.

The glasses came out about two years ago. The helmet came out a year before that and was in development for two years before that. So design decisions on which CPU to use, which RGB camera, which depth sensors were made five or six years ago. Clearly, the sensors have progressed. Since then we're still using six-year-old technology that McCarthy is stuck with right now.

It is limiting what we can do in terms of the size of the models, the field of view, the accuracy of the anchoring and the positioning. So these are all things that definitely can be improved. Comfort and compatibility are also important. So beyond just providing the AR experience, someone it wearing this new device.

We're not building AR for AR's sake, we want people to use it like any other tool in their tool belt. And for that to actually happen, they have to be able to wear it for four or five, six, maybe eight hours while they're on the site. So it has to be comfortable and it also has to fit with other things that they're using, wearing, for example, hardhats.

Current limitation, it does not work under all hardhats. There are some that you can screw down over the top. Some are designed differently, and it doesn't quite work. So you do have to be careful about what you choose.

JORDAN MOFFETT: Our fourth partner. I'll move over there. It's getting crowded over there.

DAVE TYNER: Hi everybody. How's it going. I'm Dave Tyner, thought leadership program manager for a construction customer success at Autodesk and in terms of the customer partners like DAQRI, the hardware, we're trying to enable this software accessibility of your data and your data in context through a process, which I am calling we-- there's a couple-- there's two of us, so it's we-- are calling contextual of data, which is different than visualization of data, because this is the first time somebody--

JORDAN MOFFETT: You might want that.

DAVE TYNER: OK.

JORDAN MOFFETT: I don't know [INAUDIBLE].

DAVE TYNER: Yeah, perfect. I can come over here and stand with you.

JORDAN MOFFETT: Balance the room.

DAVE TYNER: When people talk about visualization, I'm an ex visualization guy. It makes construction people, very nervous, right-- pretty pictures, little value, right. So contextualization is different from that, and it's so different that it needs a different name.

OK, so why? Why do you use BIM? Why do you care about BIM? Because BIM helps you make better decisions.

And that's whoops-- the other way. All right, BIM helps you make better decisions, right. And in a $7.2 trillion industry, global-- is that about right? Yeah-- better every decision you make comes at a time cost. And so if you can make better decisions faster, you're going to shave into that waste problem. Because there is a waste problem. I think we all know there's a waste problem, right.

And better decision-making is going to help shave that down and increase your margins. So talking about, OK, BIM. Let's just dissect it really fast. Right information at the right time to the right people to make the best and fastest decision, right.

So if we look at what does right information mean, well right information is the right data in the right context, right. And to date, this is our context. It's paper and screens, mobile or whatever paper and screens are how we're doing it. And that's awesome, and it solved a lot of problem.

But when we started really thinking about this, OK, how can we affect this, what's valuable-- VR is not a toy. It's a tool. And we can't talk about it like it's a toy. Just like when we're looking at our Revit model, we're not like, oh, this monitor is so cool. Look at the colors and everything. This is so neat, right.

No, who cares about that? As long as the technology is the center of the conversation, we're having the wrong conversation, right. So when we look at this, it's like, OK, here's real. Here's your data. Here's your people. Here's the data represented in context, and here is the physical manifestation of every decision that's being made on the job.

Now we thought, OK, cool. That works, except that the data is represented contextually on the paper. That is then sent to the human brain for translation where you're like, OK, what does that look like in real life? And all of a sudden you're running your translation program, which is robbing cycles from your decision program saying, OK, why's this look like that?

It should be about that. It should be about that, and that's what goes into reality. Am I on here? Is that-- OK, nodding heads, good. Let the record show that all 4,200 people in the room are simultaneously nodding their heads.

No, so that-- I mean, that's basically what's happening, right. And we can-- this is how we can affect this. We can affect this with augmented and virtual reality.

Granted I focus on virtual reality primarily. We are looking at augmented reality too, but it's in the adoption curve. It's back here. And what I feel is that if we can help the pre-construction people make better decisions, that's going to have a downstream positive effect while the hardware makers are making it better, while the early adopters are driving that use case so that we can deliver software-- excuse me, deliver data into the experience.

Now someone had a great question over here about how many polygons are the size of the model. That's interesting. Because when I think about that-- and my response to that how has always been when do you need to see the whole thing at one time, right. Is there a use case for that? Awesome.

Because if there is, a 3D render it seems like that would be a great tool for that. But what we're talking about is solving a problem. So if we're in this room and there's a problem that we need to solve, we don't care about the rest of the building, right. We just need to solve the problem in this room. And then when we go to the next spot, we're going to need to solve a problem in that room, which means the data needs to stream in. And then when we go over there, it's needs to stream in over there.

And this gets offloaded, and that way you maintain your cycles in the hardware. OK, I diverged. Back to this. So yeah, so how can we do this, right?

Well, the question is decision by translation. How many decisions are made in a project this size, right? About this many. About this many. And how many of those decisions-- well, you moved it. You moved it.

OK, so that's fine. That's fine. So the diagram that he showed earlier with the paper and the red marks and everything, how many decisions are being made like that, right, when you can just jump in-- is the video here?

JORDAN MOFFETT: The video is in the next--

DAVE TYNER: The videos in next slide, OK.

JORDAN MOFFETT: He's been busy, so we just.

DAVE TYNER: In a moment, we're going to show you what that piece of paper-- we're going to show you what that looks like in the immersive context, but then I'm going to head to Jordan. All right.

JORDAN MOFFETT: You keep that, but give me the clicker. There we go.

DAVE TYNER: All right.

JORDAN MOFFETT: Yeah, I moved his slides around to mess with them, since I knew he was going to be late. OK, so that's kind of, again, currently where we're at, limitations, where our heads are at, where we're trying to go. But now let's talk about what the future potential is for the value of AR in the industry.

PAUL CHEN: From the wearable perspective, obviously, it has to get much more ergonomics so someone can wear it all day and enjoy wearing it with obviously better style wise designs, better colors, newer holidays, et cetera. There are hundreds of eye wear manufacturers today, because people are very particular about what they want to wear and what they want people to be seeing them wearing.

Yes, we're making technology, but we have to be cognizant that people don't want to look like dorks when they're doing their jobs. Those are some concepts of what things might look like in the future. Nirvana, of course, is this form factor as cheap as light as ergonomic as we're used to wearing now.

We're a long ways away from that yet. Things have to get very small before we can get here. Just think of batteries, sensors, cameras, RGBs, et cetera. It's a long road.

But once we can get the better technology, we can also then provide a more immersive experience. As you saw in those videos-- those were captured actually through the smart glasses-- they're a little bit jerky and you, kind of, know you're not in the real world. You see this stuff that's floating around.

We want to make it less and less unreal and make it more real. You'll still know it's your model against the real world, but you want to make it more seamless. Here's a little AR primer. There's something called the motion to photon latency, which we're fighting to make sure that it's transparent to you.

When you move your head and you're looking at something, your perspective changes and your eyes see something from a different angle. It looks different. We're now projecting digital content that makes you think there's something there. If you move, we've got a camera that detects you moved. It sends the information from the glasses down to our computer. It tells the system the person moved we think this much and by this much angle.

The computer then has to figure out, OK, I was showing this model. Now they're over here. The model has to look like it's turned so that they think they're looking from a different perspective. It has to send that information back up to the glasses, and then the projectors have to project the content across the displays.

And we're doing this at 90 frames per second, which is 11 milliseconds. So that motion to photon latency, your motion back down and then photons onto your eyeballs, has to be less than about 10 milliseconds. That's a lot of work for us to do. You don't have to know about it.

We want to make sure that you don't even think about it. And when you move, that content moves against you and you think you're looking at something real. That's what we have to get to to make it more immersive.

JORDAN MOFFETT: But now you do know about it. So now you can quiz your buddies, right. I was joking with Paul earlier, because I was like he explained that today, and I thought, well, that just makes me feel pretty petty for complaining about the shortcomings of the technology. And then you start thinking about where we were just a few years ago. And so-- but he said it's OK.

PAUL CHEN: I'm glad to hear there are usability concerns, because that means this part is by and large solved. He's not worried about it doesn't look that real. It's real enough for now, and then they're asking for why doesn't it do this and why can't we do that. And that's why I'm here.

JOSE NEGRETE: So here's another thing we've been thinking about as we've been testing the headset is enhanced location tracking. So ideally in the future you wouldn't have to mess with a marker and try to go out in the field and have somebody lay a control point first. The headset could potentially use GPS in combination with AI to triangulate where you are in your project site and then have it load the model or the portion of the model for the area that you're standing in and then reposition as you move along rather than load the next chunk of the model that you had to section off.

JORDAN MOFFETT: And some of that's just related to the accuracy as well. Even if we do a perfect job of our initial location, you will lose depending on what software or hardware provider you talk to an eighth of an inch per 20 feet or whatever that is. So as you move further and further away from that origin point right now, you lose accuracy.

So if you're trying to do an as built validation on something that's a critical tolerance, that's really unacceptable. So the ability for the hardware to actually recognize that automatically and, kind of, always be repositioning you based on the combination of the GPS and then that finite tweaking.

AUDIENCE: [INAUDIBLE]

DAVE TYNER: So the question is if you have a single target as you move away, you lose accuracy. Can use multiple targets to sort of reposition yourself as you move? That's definitely something that we're looking at. But again, that gets to more of the worst problem of now you've got to go on site and put these targets out their.

Targets get knocked down. Even if they're lamented on paper, they might get twisted or moved. They might fall down if someone puts it up in the wrong place. So it's a point of failure. And as Jose said, if we can get to more location based or-- we're calling it mapping-- if we can build a map of your area, then no matter where you walk the system knows, oh, I know where you are and what part of the models should be there. And that will help reduce some of the accuracy issues.

JORDAN MOFFETT: So to time back really quick to the more immersive experience that Paul was talking about on the last slide, part of that is going to come with the improved field of view. So currently, the DAQRI device is at about 40 degrees horizontally and 30 vertically. So you can imagine how much of the hologram you're cutting off.

So you're seeing a-- you're seeing this reality, but you're only seeing this much of that hologram overlaid with it. So it just lends to that sense of your brain constantly knows that this isn't real. Do you have a question or you just--

AUDIENCE: [INAUDIBLE]

JORDAN MOFFETT: Just yeah, think of a question really quick. So what we want to get to is more of that 120 degrees horizontally and vertically to have that full immersive experience. So a lot of the videos were capturing now as you turn your head, that hologram is loading and loading really quickly. But it's just-- it's just something that I think we need to get over and improve upon to have full buy-in in the industry.

AUDIENCE: [INAUDIBLE]

JORDAN MOFFETT: Right. So the question is-- so the opacity-- what is the opacity on the model content?

PAUL CHEN: And currently, we're displaying it as it was done in the models. So we don't have that much control. We've heard this request. So we do have a future request into engineering for actually variable opacity. Sometimes you want to see the model as if it were solid. Sometimes you want to be able to see through it, especially if you're doing it as build versus as design comparison.

Interesting point you bring up for safety. I was down in our Forge Village booth yesterday and someone mentioned to me some job sites have a requirement if you're wearing some device like that with a hologram or other virtual content, you should only be allowed to see it when you're still. And if you're moving, they don't want any of that content showing because of exactly safety reasons. So now I have a new feature request for us to be able to detect, which we can, when you're moving, and then the hologram will turn off. When you stop, it reappears in the right position so then you see it again. Very good future request.

DAVE TYNER: The real future, which is actually the present. So yeah, so back to the VR and helping the pre-construction, sort of. Foot in the office people. So what you see here is this fully immersive, fully collaborative environment. And this is that room. This is that paper. This is the problem that they're trying to solve, which is that-- sweet usually it goes to something else. OK, good.

JORDAN MOFFETT: I'm trying to replay it.

DAVE TYNER: Oh, OK. Yeah.

JORDAN MOFFETT: You want it to keep going?

DAVE TYNER: Yeah, sure. Just replay it.

JORDAN MOFFETT: I thought I had it on loop.

DAVE TYNER: Right, so all of these stakeholders are in the experience together. They're solving the problem, right. The data was streamed in from BIM 360, no joke streamed in, no data processing in the middle through the new model coordination APIs. From there, they're free to use the plethora of tools. This is the Nvidia Holodeck, which we were down in the Nvidia booth downstairs and you should come check it out because it's pretty rad.

JORDAN MOFFETT: I'll play it again.

DAVE TYNER: OK. But it just has all these tools, just all these tools that enable people to make a good decision. It just didn't have data. And so when we saw it, I said that thing right there. We need that, and we need to pipe BIM 360 data into it and out of it.

So what they do here or we do is we're making the decision. Of course, we have measure tools and whatever. And you see me pull out that camera, and I'm taking a picture. And when I take that picture that data is sent back into BIM 360 and attaches itself to the clash as an attachment so that whoever's going to resolve it in Revit is going to bring up that attachment and say, oh, yeah, we do this, this, and this.

We can record the sessions 60 frames per second, and that goes up too. So if there is some question, it's not like I'm going to send an email. And I'm going to wait for the person to respond, and hopefully I'm on their priority respond list and I'm not 20 people deep because that's going to take them tomorrow, right. So but it's going to go right there.

I'm going to see, OK, this is what they said ah. That's the answer to the question. Boom, done, right. Time, time, time, and then cut into that-- cut into that-- increase the margins, cut into the waste problem.