tcp PointCloud Editor
AI classification - PCE
7.38 AI Classification
Important Notice
Automatic classification using AI is an experimental BETA functionality, so it may not work properly.
See licence document for more details on availability and terms of use.
Location
- Ribbon menu
- Editing
- CATEGORIES
- Classify ((BETA) Automatic Classification)
Description
This command automatically classifies the points in the point cloud using AI, and associates the points in each class with a LAS category.
Parameters
| Name | Meaning |
|---|---|
| Device | Either CPU or GPU. To be able to use GPU device, you need a graphics card that is compatible with CUDA 11.3 (more details below) |
| Model | AI model to be used. Available options: Indoor and Outdoor (10cm and 25cm) |
| Radius | A larger radius will generally result in a better and faster classification. However, if your PC does not have a powerful enough graphics card, increasing this value may cause errors in the process or slow it down considerably. If you receive errors in the classification or the process takes too long, we recommend reducing the radius value. |
| Keep existing classification | If checked, the existing classification (all classes not 0-Never classified) will be kept. If unchecked, the existing classification (all classes) will be replaced. |
| Categories | Class mapping that recognises the selected model to LAS categories. You can set the target class for each of the model classes. |
CUDA Compatibility
In order to use a GPU for running AI classification, you need a graphics device that is compatible with CUDA 11.3 or higher. Also, your GPU should support at least Compute Capability 3.7. Below are some examples of compatible GPU architectures along with sample cards for each:
-
Kepler (Compute Capability 3.7): e.g., GeForce GTX 780, Tesla K40
-
Maxwell (Compute Capability 5.x): e.g., GeForce GTX 980, Titan X (Maxwell)
-
Pascal (Compute Capability 6.x): e.g., GeForce GTX 1080, Quadro P6000
-
Volta (Compute Capability 7.0): e.g., Tesla V100
-
Turing (Compute Capability 7.5): e.g., GeForce RTX 2080, Quadro RTX 6000
-
Ampere (Compute Capability 8.x): e.g., GeForce RTX 3080, Tesla A100
For more detailed information on CUDA GPU compatibility, please refer to NVIDIA's CUDA GPUs page.
Use cases
-
Initial classification of indoor point clouds.
-
Initial classification of outdoor point clouds.