Use the Training tab in the Extract Classified Point Cloud Regions command pane to create custom point cloud classification models you can use to automatically extract new point cloud regions and/or adjust existing automatically extracted point cloud regions to your specific data. You do this by providing manually corrected classified region examples (for example, fire hydrants or electrical boxes) to train and validate the model yourself using deep-learning technology, resulting in an output model with the highest quality possible based on your input sample data. To use the custom model, simply select it in the Classification type drop-down list on the Classification tab in the Extract Classified Point Cloud Regions command pane.

Following are the basic steps for this workflow. For more detailed instructions, see Train Custom Point Cloud Classification Models.

Steps:

Commands:

1.

Ensure you have one or more LAS (.las) point cloud files containing one or more correctly classified regions (with assigned classification numbers) that represent the classes for which you want to train the new model. All files must be stored in the same folder.

Note: To successfully perform model training, this command requires an NVIDIA GPU with a minimum of 6 GB of VRAM.

2

Open the Extract Classified Point Cloud Regions command pane and click the Training tab to display the controls you will use to create ("train") a new custom point cloud classification model.

Extract Classified Point Cloud Regions

3.

Enter a name for the new model.

4.

Select the training and validation files you want to use to train the new custom model.

A training file is an LAS (.las) point cloud file containing one or more correctly classified regions that represent the classes for which you want to train the model.

A validation file, like a training file, is an LAS (.las) point cloud file containing one or more correctly classified regions that represent the classes for which you want to train the model.

5.

Optionally, change the Voxel size, a parameter that helps delineate the classified objects and appropriate surrounding context using a simple graphic view.

6.

Click the Analyze Files button to begin the file analysis process.

7.

Use the Found classes table to map one or more found classes to your new custom classes.

8.

Change Advanced Settings if necessary.

9.

When you are ready, click the Train button to start the training process.

Training and Validation Accuracy is displayed graphically during and after the training process.

10.

Optionally, click the Show TensorBoard button to display a TensorBoard in a web browser showing quality measurements and visualization during and after the training process.