Webinar | Automating territory segmentation and measurements on drone imagery

Written by
Reading time:
Data Processing & Custom Development
February 2, 2021

ATLAS for Civil Engineering - No more manual routine in construction progress monitoring. Secured data sharing with one click!

The webinar is available also in Spanish and Chinese.

What you will learn:

- Structuring your drone data for large projects
- Tracking progress on the construction site with orthomosaics and elevation
- Cut & fill volume calculations
- True DTM generation
- Converting orthomosaics into vector plan for Civil3D
- Automated object detection and territory segmentation.

ATLAS for Civil Engineering
No more manual routine in construction progress monitoring. Secured data sharing with one click! ATLAS is an AI-driven digital platform, which helps track construction progress more efficiently by automating object detection, counting and territory segmentation with AI detectors. ATLAS contains its own AI-engine, which learns from users activity and automates regular tasks in construction and maintenance.
- Timely construction reporting
- Improve teamwork and coordination
- Sensitive data is secured.


Q: How does Atlas cooperate with other AI solutions? Is it just a simple API?
A: We have our own AI engine, technically we can provide access to API but now we don't use any third-party AI solutions. But for ATLAS on premises, you can integrate your own AI model.

Q: Does it support change detection on top of the dwg of the planning layers
A: DWG file formats are not supported. Currently supported file formats are:
Images: JPEG, PNG
Orthomosaics: GeoTiff
Digital surface model (DSM): GeoTiff
Vector technical drawings in DXF
Full motion videos in MPEG-TS

Q: What about cross sections?
A: It's in our plan to release it in Q1 or Q2, we are currently designing it.

Q: Do we have access to yours, or other's deep learning datasets? A car or a tree for you will be the same for all of us...
A: We are not using our client data sets for training any detectors. We value or customer privacy and we do not share data sets of our clients.

Q: Does Atlas recognize geocoordinates of a CAD file or do you always need to manually georectify DXF files? Support for BIM data coming soon?
A: For ATLAS currently supported input data formats are PNG, JPEG, GeoTiff, DXF, KML, LAS

Q: Can we use it for industrial inspection task, with automatic annotation and reporting etc. e.g., Pix4D inspect type of workflow?
A: Yes, if you don't deal with orthomosaics but with separate images you can train detect certain objects in these images and the filter out the required image files.

Q: Do you support new Israeli grid?
A: Any EPSG non-deprecated CRS / yes

Q: I guess the question I am asking the server config being shown is this your cloud solution or just an example to run ATLAS
A: The specifications you saw in the webinar was just an example for hardware that is required to run ATLAS On-Premises. More answers regarding ATLAS On-Premises FAQ in this link: …

Q: AMD 5950 / GPU 3080 Is it enough for a server?
A: The configuration depends on the amount of data you want to store and process, but the minimal configuration may consist of 3 nodes:
Main server: Linux Ubuntu, Core i9 (6 cores), 32Gb Ram, 512Gb SSD, 1Gbit Ethernet.
Server with GPU: Linux Ubuntu, Core i9 (6 cores), 64Gb Ram, 512Gb SSD, 1Gbit Ethernet, Nvidia RTX2080Ti.
NAS: depends on the storage size
With your provided specifications it should work but again configuration depends on the amount of data you want to store and process.

Q: I have to hop off now, but can you do image by image detection, or only based on orthos?
A: ATLAS allows users to upload separate Image files and Geotiff maps generated in other software (like Pix4D, Metashape or some other). Worth mentioned is that ATLAS can build maps from raw images as well.

Q: Is it possible with a thermal camera, identify different areas and calculate them using AI in Atlas?
A: Yes, if you work with JPEG, if use Geotiff files then not yet but we are planning to add such feature for this format as well.

Q: Are you taking steps to market to the US government?  DJI drones are on a ban list.
A: In case of ATLAS you can use any type of drone or use satellite images an upload to the ATLAS. In case of UgCS then there are other options to choose and have a flight with other drones and we are not locked on DJI drones only

Q: What type of volumetric measurement do you have or plan? (Smart, Bench, previous survey, plan....)
A: Right know we do simple volume calculations for specified areas and in our roadmap, there are multiple things for cross-sectional diagrams and 3 surfaces.

Q: Are there any integrations with the ArcGIS platform, we are specifically interested in how to leverage the Atlas AI/ML output in ArcGIS Online?
A: ATLAS outputs json files and other vector formats that can be used in ArcgisOnline, currently there isn’t such a feature available with a single push of the button, but there are multiple different integrations and simplifications in our roadmap.

Q: What about reality capture?
A: ATLAS has a scalable architecture. So, any photogrammetry engine that support command like interface our API can be integrated in ATLAS

Q: Does it support the iPad/iPhone operating system?
A: ATLAS is web-based app and user interface is just like and usual website but for small screens we reduced number of available features due to its it impossible to do specific complex features on smaller screens with just only fingers.

Q: Is it suitable for creating crop classification train data as well?
A: Yes, you can train multiple detectors to detect specific crop.

Q: Do you plan on expanding the server capability with only 12 cores I can see large projects taken extended processing?
A: We are not limited in number of cores. We can utilize any cores you have. But in general, you can several PC to increase the performance etc.

Q: Can the AI detect plant type like mango tree /leaves in an environment like the picterra?
A: General answer is Yes, just pay attention to GSD

Q: Is Atlas hosted on Amazon, Azure, or some other cloud hosting platform?
A: We are now using Amazon

Q: Can we specify many areas with vegetation and interpolate all?
A: Yes, you can

Q ATLAS is running online, or you need to install in the computer?
A: ATLAS runs online, but there is an option to have it installed in companies closed infrastructure if necessary, for security.

Q: And do you support now Mavic Air 2?
A: For ATLAS you can upload data from any drone type. The mission planning and flight control software UgCS doesn't support Mavic Air 2. Screen reader support enabled.

Q: Are you supporting Autel Evo 2 Pro RTK?
A: For ATLAS you can upload data from any drone type. The mission planning and flight control software UgCS now has the option to make mission export in .kml format compatible with Autel Evo 2.
can we use Atlas in a system out of internet?
Atlas on Premises can be used without internet

Q: Are you supporting Matrice 300 RTK?
A: For ATLAS you can upload data from any drone type. Regarding the mission planning and flight control software UgCS. Recently we have released UgCS for DJI (2.28.87) BETA version for Android devices which supports M300. It's now available on our website. Please use it with CAUTION!

Q: For example, how many datasets have you processed to train the ai how much is human work is done on the data sets?
A: It depends on the object you will analyze and on the final quality of the analysis process. You can check out a few user cases we have published at our website for additional background of process and how much time is required for final results.