Auto Classification
This command classifies point clouds automatically.
Requirements
Select one or several clouds and launch the command. Clouds will be processed independently unless they are merged before.
Select an appropriate Classification model. A table displays which Categories can be detected by the current model and for each one in which Class the points will be added.
If necessary, modify the target classes thanks to the button Edit category mapping. It will open Classification Settings.
If necessary, use Manual Classification to improve the classification.
Technical information
The classification process is based on machine learning algorithm that relies on GPU computation.
For optimization purpose, the installer doesn't include all the required elements. Most of them are stored online and are automatically downloaded and installed when needed.
Checking requirements
At startup, the software checks if all the requirements are fulfilled. If it is not the case, a warning is displayed on the icon of the command and a form replaces the command, allowing to download and install the missing elements.
Here is the list of the requirements:
Your computer needs to have a Nvidia graphics card with GPU compute capability. Check the list here: https://developer.nvidia.com/cuda-gpus. A good GPU is advised to get acceptable computation time.
The correct Nvidia CUDA Toolkit needs to be installed on your machine.
The classification library.
At least one classification model.
The classification library and the models are downloaded and recorded in the Classification folder defined in the Classification Settings.
Downloading requirements
Once opened, the form starts to download and install the CUDA Toolkit and/or the Classification library.
At this step, if no models are available in the Classification folder, all the available models (only the latest versions) will be downloaded as well.
The software will restart once the setup is done.
New models and updates
The training of classification models is an ongoing process. Therefore, model updates are regularly released, as well as new classification models.
At startup, the command will perform a check to detect updates and invite you to download them if you want.
Troubleshooting
It is also possible to install manually:
the CUDA Toolkit. The current classification engine requires CUDA Toolkit 11.2.0. It can be downloaded here: https://developer.nvidia.com/cuda-toolkit-archive.
the classification library. Download the classification offline package available here and follow pdf instructions.