Skip to content

Data processing

Thanks to the digital twin and our AI, the extraction of inventory data is done at more than 200 km per hour

Our data handling process

At Jakarto, we capture HD geographic data that we process and then broadcast in 3D. With this data, we create what we call a digital twin that will represent a specific territory. This is achieved by capturing data from our mobile mapping units. Our sensors allow us to obtain a georeferenced 3D point cloud.

Jakarto then uses artificial intelligence to transform the raw data into intelligent, usable, and ready-to-use data within a few hours.


Mastering and using artificial intelligence

Artificial intelligence consists of algorithms that can find the answer to questions instead of an expert.

At Jakarto, we use AI to detect a specific object in a geospatial dataset in order to map it.

The advantage? It's a huge time and money saver because the process is automated. In addition, the machine does the work at all times, which allows the human to concentrate on more important tasks.


Object recognition in AI

  1. Concretely, we select the points of the object in question and we tell our algorithm that they represent this object.

  2. From the segmented 3D data, we can create geospatial data that our clients can use. They can then benefit from a geolocalized inventory created automatically. We use supervised algorithms for this: they are based on the knowledge that Jakarto's experts give them.
25% 3D data
25% 3D data
50% Segmented 3D data
50% Segmented 3D data
75% Object recognition
75% Object recognition
100% Machine learning results
100% Machine learning results

How does Jakarto use Artificial Intelligence?

A work which is done in 4 steps and repeat!

Step 1 - Annotation

Annotate the data Jakarto_Annote_200x200_W

Virtual scouting work

  • Cross-referencing with an existing database or in-house creation.

Point cloud

  • Our experts manually recognize the objects and select the points of the object.
Step 2 - Training

The algorithm Jakarto_Algorithme_200x200_W_V02

Training phase

  • Algorithm that learns from data
  • We obtain a trained model
  • We can take an existing model to refine it and adapt it to our objectives: it becomes tailor-made for our problems

Data science / deep learning / machine learning

  • Know the state of the art to exploit the most efficient algorithms
  • Work in collaboration with researchers to create new algorithms
  • Generate models according to the tasks to be solved
Step 3 - Automated Detection

Detection Jakarto_Detection_200x200_W

Application of the model

  • Application of the model on the data set
  • Conversion of results into simplified data formats (geographic layers)
  • Delivery of results to business experts

Business application

  • Automatic inventory delivery
Step 4 - Validation

Validationoutput-onlinepngtools (9)

Data verification

  • Data extraction, validation and correction using algorithms
  • Delivery of results to business experts

Business application

  • The data is ready to use for our customers

Vous avez un projet d'acquisition de données géographiques en 3D?

Planifier une démo