Classify by Deep Learning

Functional Overview

This function uses a deep learning segmentation model to classify point cloud data, suitable for uav (unmanned aerial vehicle) data, mountainous or rural scenes. Now supports 2 categories: vegetation, buildings. Developed based on a sparse convolutional model, the algorithm ensures high efficiency while bringing better detail performance.

Usage

Click onClassification > Classify by Deep Learning

Parameters Settings

• Data to be processed: Select single or multiple data loaded in the software.
• From Class: You can select the category that needs to be classified, and the unselected category will not be covered by the model classification result, which is especially effective when there are some fine categories in the data.
• Sence:Choose Urban or Rural sence, to use different deep learning model.
• Classify xx: Check to enable this category, you can select the number corresponding to this category.
• Height Above Ground: Select the ground point of category 2 as the reference ground point and fill in the ground height. This parameter can effectively prevent the ground point from being wrongly classified as a building.
• Use GPU:Using GPU acceleration to improve the performance of classification.

  Note: The output of this function will overwrite the original data file, users who need it should back up the data by themselves. Due to limited training data scenarios, deep learning models may perform poorly on certain scenarios and certain types of data.