Our activities

We share our knowledge and experience with others. We are open to organise trainings related to geoinformation and remote sensing. The training materials are dedicated and customised to specific recipients/customers.

How is machine learning used in geoinformatics?

OBJECTIVE:

Presentation of the methods, technologies and ways of application of machine learning in geoinformatics. Delivering a lecture and conducting practical exercises divided into two main parts, the former of which was dedicated to geodata characteristics and operation of machine learning algorithms, whereas the latter one was associated with the introduction to the technologies of deep machine learning.

RESULT:

The approach applied in machine learning to solve spatial-related problems was presented within the framework of the theoretical part. The data used in geoinformatics based on the examples of the projects previously executed by the company was thoroughly described. The random forest algorithm was explained. The methods of statistical evaluation of model efficiency were presented.

 

The practical part of the workshop session involved presenting the method of point cloud classification by means of extracting geometric features by the random forest algorithm and the method of classifying aerial images by means of AdaBoost algorithm. Subsequently, particular components of perceptron and convolutional neural network were depicted. The working principle of the issues related to operation of neural networks, such as the backpropagation algorithm, optimisation algorithms and function convolution, was described. Additionally, the basic techniques which support teaching (deployment) of deep machine learning models were presented. The last phase of the theoretical part consisted in presenting the most popular neural network architectures.

 

The classification of the point cloud by a convolutional network featuring a perceptron layer into a fully connected network, as well as classification of aerial images by means of U-Net architecture convolutional network were carried out within the framework of the practical part of the second workshop session.

CUSTOMER:

Training provided within the framework of the FabSpace 2.0 Project

OPEGIEKA Sp. z o.o.

Click here for directions.

Head Office

Aleja Tysiąclecia 11

82-300 Elbląg

phone: +48 (55) 237 60 00

fax: +48 (55) 237 60 01

E-mail address:

lab@opegieka.pl

Branch in Warsaw

ul. Grzybowska 80/82 lok. 700

00-844 Warszawa