Training models

Traning point cloud classification

Collect point cloud dataset as describe in 3D dataset collection, and preprocess them as explained in 3D dataset preprocessing.

To train the models, you need to clone this repo, and install the requirements to train the model.

Training deep learning model

The supported models are available under config/config.yaml.

#python trainer.py --model <model_name> --train
#an example for training DGCNN with color (DGCNNC)

python trainer.py --model DGCNNC --train

Adding new model

ToDo

Training 2D object detection

ToDo

Note

You can train the model on H-BRS Scientific Computing Cluster, provided that you have access to it. The tutorial on how to submit job and train your model on the cluster can be found here.