Custom Tree Species Classification Model Training
Function Overview
Custom Tree Species Classification Model Training trains a deep learning model for tree species recognition from user datasets. The workflow is similar to Custom Deep Learning Classification, but the training data and model are used for tree species recognition.
The trained model can be used later to define a tree species recognition tool.
Usage
Open the custom tree species recognition model training function in the TLS Forest module.
The workflow is basically the same as Custom Deep Learning Classification, but some parameters and dataset requirements are different.
Add Model Training Task
Both point cloud mode training and panoramic mode training are supported.
- Point Cloud Mode Training: The training data and validation data must meet the requirements shown in the prompt. Each point cloud must have been segmented into individual trees and must contain the TreeID attribute. Each point cloud also requires a CSV file with the same file name. The CSV file contains two rows and specifies the tree species name corresponding to each TreeID.
- Panoramic Mode Training: The training data and validation data must include image files and the corresponding tree species label files (CSV), which contain image names and tree species columns.
The following figure shows an example of the training data:
