Jakarto's mastering high-definition 3D data processing

Jakarto applies a series of sophisticated processing techniques to the raw data captured by its sensors. It’s how the platform generates data that’s rich, ready-to-use, verifiable, and precise.

And when it comes to generating inventory elements automatically or semi-automatically, artificial intelligence ensures the Jakarto platform performs above and beyond.

Jakarto Data

Data usage

Data processing

Data acquisition

How does Jakarto process data?

Jakarto collects HD geographic data, processes it, and renders it in 3D.

It creates what’s called a digital twin, a virtual replica of a specific city or landscape. The technology is made possible thanks to Jakarto’s mobile mapping units, which capture the data.

Using its cameras and LiDARS, Jakarto creates georeferenced imagery and 3D point clouds.

Next, it uses AI-based processing methods to transform raw data into intelligent, ready-to-use data.

Why artificial intelligence?

Artificial intelligence consists of algorithms that, when well designed and trained, answer their own questions instead of deferring to experts.

Jakarto uses AI to detect specific objects in its geospatial datasets and map them.

The AI advantage? It saves considerable time and money. The process is automated, and the algorithm runs continually, letting us humans focus on more important tasks.

Jakarto’s developments in AI

The 3D data that we capture represent reality at an instant T and are translated by point clouds.

Once collected, Jakarto’s 3D data is translated into point cloud form, representing a snapshot of reality. Next, thanks to an algorithm, objects and their positions can be identified with precision.

How do you customize an algorithm at Jakarto?

How do you identify objects?

Simply put, Jakarto identifies points outlining an object then tells its algorithm what these points represent.
From the segmented 3D data, Jakarto creates geospatial data, giving its customers access to an automatically generated and geolocated inventory of objects. For this process, it uses supervised learning algorithms, which work off knowledge provided by Jakarto’s experts.

3D data

25%

Segmented 3D data

50%

Object recognition

75%

Machine learning results

100%

See an algorithm in action

How does Jakarto use Artificial Intelligence?

A 3-stage process

Step 1

Annotation

Virtual surveillance work

Jakarto cross-references its data with a pre-existing or in-house database.

Point clouds

Its experts manually recognize objects and create point clouds for them.

Step 2

The algorithm

Training phase

The algorithm learns from the data
Jakarto creates a trained model
Its experts use existing models to refine the algorithm and customize it to your needs

Data science / deep learning / machine learning

Jakarto uses state-of-the-art methods to create more efficient algorithms
It collaborates with researchers to create new algorithms
It generates models customized for various tasks

Step 3

Detection

Applying the model

It applies the model to the data
It converts results into simplified data formats (i.e., geographic layers)
It delivers the finished data to business professionals

Business applications

The automatic inventory is delivered
The data is ready for customer use

The 5 steps of teaching object recognition to an algorithm

Data preparation phase
Jakarto provides the algorithm with raw 3D data samples along with results indicating what these samples are meant to represent.

The more annotated examples given to the machine, the more it learns how to identify the common characteristics that separate one object from another.

The training phase
Next, Jakarto uses a computational graph to attempt to categorize the 3D data. Its team compares the model’s results to the expected outcomes and shows both datasets to the algorithm.

By seeing its own errors, the algorithm learns how to self-correct and improve.