Model2Cloud

Generate point clouds based on input 3D building model.
The code repository for manuscript "Automatic Generation of 3D roof TRaining Dataset for Building Roof Segmentation from ALS Point Clouds"

Usage

1. Generate clouds

1.1 Generate from single .obj file

use tools/model2cloud.py to achieve the generation of point clouds.

$ python3 -m tools.model2cloud --models_dir test_data/ --clouds_dir test_data/test_out/ --sample_mode point_density --pt_param 20 --noise_sigma 0.02

parameters

$ python3 -m tools.model2cloud --help
usage: Model2Cloud [-h] --models_dir MODELS_DIR [--clouds_dir CLOUDS_DIR] [--sample_mode [{point_density,point_spacing}]]
                   [--pt_param PT_PARAM] [--noise_sigma NOISE_SIGMA]

optional arguments:
  -h, --help            show this help message and exit
  --models_dir MODELS_DIR
                        the path or directory for models need to be processed.
  --clouds_dir CLOUDS_DIR
                        the directory for saving generated clouds. if no input is received, the clouds wont be saved.
  --sample_mode [{point_density,point_spacing}]
                        the sampling mode: whether the pt_param is point_density or point_spacing
  --pt_param PT_PARAM   the parameter for sampling
  --noise_sigma NOISE_SIGMA
                        the sigma value of additional guassian noise (created by normal distribution)for generated point
                        clouds

1.2 Generate from .citygml file

use tools/model2cloud_citygml.py to achieve the generation of point clouds for a series of 3D buildings in a region.

$ python3 -m tools.model2cloud_citygml --models_dir test_data/gml_files/Altenbach --clouds_dir test_data/gml_objs/Altenbach --t_rm_abnormal_face 0.3 --sample_mode point_density --pt_param 20 --noise_sigma 0.02

parameters are the same as 1.1 parameters

2. cloud-cloud distance to evaluate the quality of generated clouds

use tools/eval_dist_c2c.py to achieve the cloud-cloud distance evaluation.

$ python3 -m tools.eval_dist_c2c --comp_dir test_data/test_out/clouds/ --ref_dir test_data/ --comp_list_path test_data/test_comp_list.txt --save_res True

paramters

$ python3 -m tools.eval_dist_c2c --help
usage: Model2Cloud [-h] [--comp_dir COMP_DIR] [--ref_dir REF_DIR] [--comp_list_path COMP_LIST_PATH] [--log_dir LOG_DIR]
                   [--save_res SAVE_RES]

optional arguments:
  -h, --help            show this help message and exit
  --comp_dir COMP_DIR   the directory for generated clouds. required.
  --ref_dir REF_DIR     the directory for reference clouds. required.
  --comp_list_path COMP_LIST_PATH
                        the path of list including the point clouds names for comparison and ref. list example:
                        ['1_com.txt 1_ref.las', '2_com.txt 2_ref.las', ...] each path in the list will be organized as
                        'comp_dir'+'list[i]'. default=None. If None is recieved, all '.txt' file in comp_dir will be
                        considered.
  --log_dir LOG_DIR     the directory for logger. if none, use the upper directory of comp_dir as the logger dir.
  --save_res SAVE_RES   whether save the eval result to file. If True, the eval result will be saved as .csv file in the
                        log_dir.

Requirements

The following requirements are necessary for Model2Cloud:

• laspy
• triangle

Other potential requirements can be found in requriements_m2c.txt and according to the compilation errors.