# Generating Fully Automated 3D Site Maps with Zero Touch, Every Flight - Event: NestGen ’26 - Published: October 2, 2026 - Length: 21:44 - Watch: https://nestgen.org/library/watch/XPpatigC5JI/ > The transcript of a recorded NestGen session, for reading or for use with an AI tool. It comes from the video's > captions; automatic captions can mishear names and terms. ## About this session Keeping site maps current usually means planning survey flights, collecting the data by hand, and waiting on processing before anyone can use the result. By the time a map is ready, the site has often already changed. In this NestGen session, FlytBase will show how Akara generates fully automated 3D site maps from every flight, with zero manual touchpoints, and what teams should think through when moving from occasional surveys to maps that stay up to date on their own. ## Transcript **[0:00]** more. The limit now is how fast the map gets made. Let's pull numbers on it. A typical mapping flight for any site uh is about, you know, 500 images. Three flights a day is about 1,500. A month is 145,000 images. Sorry. Uh sorry, it's about 45,000 images. Now from one dock on one site, multiply that by number of docs you're planning to deploy. Not one of those images is useful until it becomes a map. An image is data. A map is what you make a decision on. And it isn't only volume. Every one of those flights has to be checked, stitched, aligned to the last one, and put where the team can find it. With one flight a month, a person **[1:01]** becomes a full-time job. And usually it's a job nobody was, you know, hired for. And here's the more uh uncomfortable part. Two years ago, general purpose mapping softwares were were good enough. You had one pilot, one flight a week, and someone at a desk to process it. At 50 docs running automated workflows, it isn't. This is where it breaks. If you ask me what the dumbest thing about mapping is today, it's this. Most mapping softwares was built for a pilot with an SD card. It was never designed for a dock. It falls over in five places. One, it's built for one flight at a time. Upload a job, wait and export. That works once a month. A dog lands every few hours every single day. Two, docs hand over data their own way. Flights sync on the doc's schedule, not when a person walks up with a card and the tools are **[2:01]** still waiting for the person. Three, the base map is always a project. Someone commissions it, someone makes it, it gets delivered and it's out of date the next morning because the site is moving every day. Operations don't need last quarter's export. They need today's export. Number four, the connection is one of the bottlenecks. These sites run on LTE, on Starlink, on fiber, and if you're lucky, every upload completes with the next flight for the same link. And five, the one that actually kills programs, a person is still in the middle. Someone downloads the images, starts the job, checks it, exports it, shares it every flight. We automated the flying and kept all the paperwork. When a drone program fails, uh it's rarely because the drone didn't fly. It fails on gap five, the week that person goes on leave. So every day you just keep falling, you know, behind. And here is what that looks **[3:05]** like over a week. Top row you can see flights land every day, 1500 images each. Bottom row you can see maps get made, someone uh gets to them. Monday's imagery, then Tuesdays. By Sunday, you're still working through the start of the week. So, you've bought a drone that gives you daily data and you're still making monthly decisions. The problem isn't the drone or the camera. It's the part between the landing and the decision. And this isn't uh a small problem. McKenzie found that large projects typically run about 20% behind schedule. uh Autodesk and FMI estimate that bad uh you know uh data is behind around 14% of all rework in construction. Those numbers are from construction alone but anyone running a mine or a plant knows the same rule. When the picture is wrong uh the work is wrong. So now let's just step back for a second. uh every morning whoever runs a site uh answering uh you know **[4:07]** uh four questions. Where do the crews go today? How much material is left? What changed since yesterday? Is anything unsafe or out of place? Every one of those answers come from a map, either a real one or in someone's head. And the answer is the only as good as the map is fresh. You can't run a site daytoday off a picture from the last month. Picture a site manager at 6:00 in the morning, coffee in one hand, uh phone in the other deciding where 40 people go today. Right now that decision gets made on a map from the last month. So that is what we are trying to solve here uh you know with something I will be introducing you with. Uh so picture the same site uh mapped every day. Walk down this table with me. uh who flies not a person driving out a talk on schedule. How often? Every single day. You know, flight to finish map the same day again. People touching the data. Uh not two or three. None. No. No. No one's touching the data. And the row **[5:12]** that matters the most. Decisions get made on today's map. You know, that's the whole idea. So now uh notice what's not on this table. A better camera, a faster drone, a smarter model. None of that changes the answer to the frequency or to the first question I asked you. Frequency does. That's the vision every decision on a site made on a map less than a day old. The question is how you get there without hiring a processing team to sit behind every dock. So that's why we built Aara. It didn't start as a product idea. We watched customers put docks on site and get stuck uh in exactly this gap. One of them uh a solar builder had their own engineers build a processing pipeline just to have a fresh map every morning. That told us two things. Uh first of all, the need is real and most companies will never build that themselves. So Aara closes the whole loop for you. You set the schedule once, the doc flies, the images upload on their own. Accara builds the map, the map lands on your site on **[6:15]** your FlytBase operations dashboard. Your team makes the call and next flight is already booked. Nobody touches an SD card, starts a job or exports a file. If you remember one line from today, make it that one. Here is what it looks like step by step. Six steps and you only touch the first two basically. So, one, plan the mission uh once in flight base, the same place your dock already flies from. Two, import it into AAR and set it to repeat. That's the last time a person is involved. Three, the dock launches the drone uh on schedule. Four, the drone captures the data and it uploads to the cloud on its own. Five, processing starts on its own. Nobody opens a job. Nobody presses go. And six, the finished map appears back in FlytBase as a live layer right where your team is already looking. That's why we call it native. Accara isn't **[7:19]** a separate tool you export into. No new login, no file transfer, no handoff. Set it once and everything after that is automated. So rather than uh you know just describe it, let me show it to you. Uh yeah, just a second. I hope my screen is visible. Uh, and you guys can hear it. >> A construction site changes with every passing day as the landscape is constantly reshaped. So do mines, forestries, and stockpile yards where the terrain is in a constant state of change. But every decision on that site depends on knowing its its exact shape today. But capturing it is still manual. Fly the site, pull the SD card, upload the images, wait for an external tool to process them. Hundreds of steps for every map. So teams capture occasionally and the map is always weeks behind the site. Introducing Aara. It removes the manual loop. Plan the mapping mission once in FlytBase. **[8:23]** Import it into AAR and set it to repeat. The doc launches the drone on schedule, captures the data, and uploads it to the cloud automatically, ready for processing. No site visit, no SD card, no manual upload. Once the flight lands, processing starts on its own. Aara intelligently stitches every captured image into a seamless orthomosaic, a single corrected map of the whole site where every pixel sits in its true position. Accara also builds the digital surface model and the digital terrain model. Navigate the site through an interactive 3D mesh. Inspect every detail within the point cloud and generate accurate elevation contours across the terrain. Measure distances and areas straight on the map. Draw a boundary around a stockpile or a cut and Aara computes the volume from the terrain model. Compare any two dates and see exactly what moved, what was built, and how much was hauled. Every result flows back as a layer on the flight-based site map, so the people running operations are looking at current ground truth, not last month's. Export it for your records or share the map with your team. Map your site as often as it changes with a car. **[9:39]** Awesome. So basic basically uh I hope uh that answers a lot of your questions uh and you know helps you and your uh team understand what AARA is doing uh and what all gaps is it filling everything uh from flights to data collection to processing and then to uploading it to the base map is completely you uh end to end automated. So uh what you saw again like it closes a few of the gaps uh over here. So today it's a project uh with AARA every single flight uh is processed and sent to the map layer on its own. So Monday's flight uh will go on Monday, Tuesday's flight will go on Tuesday and you know basically so on and that is how it works like someone has to start a job uh these are the gaps that it is closing out for you you **[10:43]** know uh nobody uh keep that up nobody has to you know basically start the job uh after a flight uh you just set the schedule once uh three flights a day for a week 21 flights 21 maps uh zero jobs you know started by hand and every map is compared with the one before. So what you open in the morning isn't just a side uh it's what changed on the side. Uh that compare view is where most of the value shows up. Nobody wants to study a full map every morning. They want to know what's different. Uh a stockpile that dropped, a trench that opened, uh equipment that moved, a road that got built. So every single detail can be you know viewed very easily. uh and the gap number four uh the connection. So this is where uh we're taking it like the data that you're building the the data that you're getting uh it moves in the background and it gets processed in the cloud and the dog keeps flying. So what uh **[11:43]** my customers have have told me Starlink uh does not uh like because of bandwidth you are not able to basically uh produce uh data or upload it properly you know every single day like it's not that easy uh because if the doc uh internal storage gets full or something like that happens then it's a bottleneck for everyone and uh on a remote site it just matters more than it sounds. So that is why you know the data also becomes a bottleneck. Uh at Akara we try to complete a fast mapping uh which basically is able to give you insights on a better scale. uh you don't need like a very high quality 3D mesh or a highquality 3D map or an ortho for everything for the insights that you need a fast map uh built quickly on uh data can help answer all your questions. So there are certain ways uh in which uh you can **[12:48]** basically know your site and uh here's what uh usually comes out of one of those flights. So there's uh terrain uh models that you can make. You can make 3D models uh you can move around in and then there is volumes you can measure without walking a pile and you can compare between two dates uh the same outputs you would expect from a survey. So similar type of things. Uh the difference is you get them every day without asking for them. And because they come from the same dog uh on the same schedule or the same area, they line up. So day 10 compares cleanly with day one. Uh and that consistency is what makes trends uh worth trusting. How fast a pile is moving, how many rows went in this week, whether a slope is shifting or not. So that is where uh it's very helpful. Uh and uh none of this came out of a lab. Uh it runs every single day uh on real sites. A solar builder uh maps it site every **[13:53]** morning with docks and that morning map decides where the crews go next. Uh that's exactly the behavior uh we design Aara around. What uh struck uh me about that team is that they never asked us for mapping. They asked for a way to know every morning which blocks were ready for the next crew. The map was just how they answered it. [clears throat] That's the lesson for all of us. Like nobody, you know, buys maps. They usually buy decisions. So, as I mentioned before, instead of making those insanely highquality maps uh which you don't need, you can make a fast map and get the same level of insights and make better decisions on it. uh another customer of ours like a mining operation uh that runs around uh you know like uh quite a few maps a day. Uh stock piles and pits are measured from dock not by a survey crew uh walking the pile and one of our partners in Europe uh did what every serious operator should do uh before putting their name on a single map. they benchmark Accara's accuracy against the tool that they already **[14:57]** trusted and I'd encourage all of you to do the same like we would definitely you know want you to test it then take my word for it so all of you who have a doc who have a FlytBase account are more than uh happy to you know uh test it out and basically you know learn and give us feedback on how to make it or make that experience better for you. So now I want to highlight and I want to tell you uh basically what broke along the way because that's more useful than you know another simple feature slide that we can make. So every one of these came from customer conversations between April to July this year. Uh we didn't find them in testing. Uh customers found them for us by using it every single day. So daily sites turned out bigger than anyone planned for and d sites hit the image limit on a single map. So one of the lessons that we learned was plan for the site you'll have in a year not the one that you have today because the site just keeps on **[15:59]** increasing uh if if you're if it's not a closed one. Uh then maps were processed one at a time when you map daily a queue turns into a backlog. So we started doing par mapping for most of our customers. Then some flights came back thin with few images and uh the quality of maps were not that good. So the way you make missions is also very important. Uh and uh the first question that people asked were how accurate this is. So you know it follows the same uh like accuracy as uh DJI Terra as drone deploy and is exactly at that same level. Uh we have benchmarked it against all of the industry standard mapping softwares as well. So now uh just to close this out uh you know what do you do on a Monday when someone has to go on a construction site. So if you're a partner uh stop **[17:00]** quoting docs as flights quote every doc with a daily map uh the doc is the hardware the daily map is what your customer is actually buying. Uh then just you know show the customer then and if you are an enterprise pick one site map it daily for every 30 days uh count the decisions and then you will be able to see the value of data analysis on the decision making on how it has changed and how you are not facing delays and how the decisions that you're making are just uh better for every single day. So yeah uh first of all I would like to thank everyone uh who joined this uh session with us. Uh docs have definitely made uh the flying automatic. Uh the next step for us uh for us all is how do we make the map making automatic. So everything uh people are excited about after that the alerts the AI the automation starts with a map that is fresh. So um I leave you where we started. **[18:01]** Uh how old is the map that you are making your decisions on and if the answer is anything other than this morning then you can just drop me a note in the chat uh and we can you know have like a session we can grab a session together uh after this. So yeah uh thank you so much everyone uh again for your time on this and I would like to now take uh uh a few questions. So let me just see uh yeah so first one says I'm working on a fixed wing US project 2 kg. How can FlytBase help me if I can integrate with a docking station? Uh it's not related to a map but uh we definitely you know just to give you some context we definitely have uh like a software stack uh which is compatible with uh all the cloud APIs of uh docking **[19:05]** stations and drones of different manufacturers and then we can also integrate with uh Mingling protocols PX4s and autopilots um and your cubes as well. So overall for mission planning for data collection we can help you with that just to you know answer that question. Um what is the cost of uh generating these maps? Uh is the second question and to be honest it is quite less uh the only uh than what you are currently paying for and we have calculated that in the back end. It depends on how fast you want it and what type of GPU or server specs you're using. But overall the cost of building the whole map like let's say a,000 uh you know a,000 image map uh every single day is less than $5. How is this different from uh drone deploy? So drone deploy is as of this moment **[20:10]** uh a software which is solely used for making digital twins. Uh how it started uh was a drone first software. So it's not a doc first software. Uh FlytBase is a doc first software and Aara is one part of it. Uh your site does not only need maps. It also needs security. It also needs other parallel you know threads going on. So FlytBase is a platform for your fleet operations and drone autonomy and then you can uh you know basically you're going to have a lot more uh [clears throat] things uh going on uh parallelly. So once you make the map uh in the morning in that day now if you want you can do security at night. So FlytBase offers you the full stack for that. Uh but yeah so I believe these were all the questions. If uh anyone has uh any other question uh you can just ask them uh in the Q&A tab. Uh but yeah uh thank you so **[21:15]** much uh everyone for joining in and lovely uh having all of you. Uh there are a lot of more sessions that we have going on uh at NestGen. So please uh go and join those sessions, learn from people's experiences and yeah build something great. Thank you so much. Have a good day.