import glob from pathlib import Path import logging import numpy as np import open3d as o3d from ..utils import DATASET from .base_dataset import BaseDataset, BaseDatasetSplit log = logging.getLogger(__name__) class TUMFacade(BaseDataset): def __init__(self, dataset_path, info_path=None, name='TUM_Facade', cache_dir='./logs/cache', use_cache=False, use_global=False, **kwargs): """Dataset classes for the TUM-Facade dataset. Semantic segmentation annotations over TUM-MLS-2016 point cloud data. Website: https://mediatum.ub.tum.de/node?id=1636761 Code: https://github.com/OloOcki/tum-facade Download: - Original: https://dataserv.ub.tum.de/index.php/s/m1636761.003 - Processed: https://tumde-my.sharepoint.com/:f:/g/personal/olaf_wysocki_tum_de/EjA8B_KGDyFEulRzmq-CG1QBBL4dZ7z5PoHeI8zMD0JxIQ?e=9MrMcl Data License: CC BY-NC-SA 4.0 Citation: - Paper: Wysocki, O. and Hoegner, L. and Stilla, U., TUM-FAÇADE: Reviewing and enriching point cloud benchmarks for façade segmentation, ISPRS 2022 - Dataset: Wysocki, Olaf and Tan, Yue and Zhang, Jiarui and Stilla, Uwe, TUM-FACADE dataset, TU Munich, 2023 README file from processed dataset website: The dataset split is provided in the following folder structure -->tum-facade -->pointclouds -->annotatedGlobalCRS -->test_files -->training_files -->validation_files -->annotatedLocalCRS -->test_files -->training_files -->validation_file The indivisual point clouds are compressed as .7z files and are stored in the .pcd format. To make use of the dataset split in open3D-ML, all the point cloud files have to be unpacked with 7Zip. The folder structure itself must not be modified, else the reading functionalities in open3D-ML are not going to work. As a path to the dataset, the path to the 'tum-facade' folder must be set. The dataset is split in the following way (10.08.2023): Testing : Building Nr. 23 Training : Buildings Nr. 57, Nr.58, Nr. 60 Validation : Buildings Nr. 22, Nr.59, Nr. 62, Nr. 81 Initialize the function by passing the dataset and other details. Args: dataset_path: The path to the dataset to use. info_path: The path to the file that includes information about the dataset. This is default to dataset path if nothing is provided. name: The name of the dataset (TUM_Facade in this case). cache_dir: The directory where the cache is stored. use_cache: Indicates if the dataset should be cached. use_global: Inidcates if the dataset should be used in a local or the global CRS Returns: class: The corresponding class. """ super().__init__( dataset_path=dataset_path, info_path=info_path, name=name, cache_dir=cache_dir, use_cache=use_cache, use_global=use_global, # Diese habe ich selbst hinzugefügt **kwargs) cfg = self.cfg self.name = cfg.name self.dataset_path = cfg.dataset_path self.label_to_names = self.get_label_to_names() self.use_global = cfg.use_global if info_path is None: self.info_path = dataset_path if self.use_global: # Finding all the training files self.trainFiles = glob.glob( str( Path(cfg.dataset_path) / 'pointclouds' / 'annotatedGlobalCRS' / 'training_files' / '*.pcd')) # Finding all the validation Files self.valFiles = glob.glob( str( Path(cfg.dataset_path) / 'pointclouds' / 'annotatedGlobalCRS' / 'validation_files' / '*.pcd')) # Finding all the test files self.testFiles = glob.glob( str( Path(cfg.dataset_path) / 'pointclouds' / 'annotatedGlobalCRS' / 'test_files' / '*.pcd')) elif not self.use_global: # Finding all the training files self.trainFiles = glob.glob( str( Path(cfg.dataset_path) / 'pointclouds' / 'annotatedLocalCRS' / 'training_files' / '*.pcd')) # Finding all the validation Files self.valFiles = glob.glob( str( Path(cfg.dataset_path) / 'pointclouds' / 'annotatedLocalCRS' / 'validation_files' / '*.pcd')) # Finding all the test files self.testFiles = glob.glob( str( Path(cfg.dataset_path) / 'pointclouds' / 'annotatedLocalCRS' / 'test_files' / '*.pcd')) else: raise ValueError( "Invalid specification! use_global must either be True or False!" ) @staticmethod def get_label_to_names(): # """Returns a label to names dictionary object. Returns: A dict where keys are label numbers and values are the corresponding names. """ label_to_names = { 0: 'not_assigned', 1: 'wall', 2: 'window', 3: 'door', 4: 'balcony', 5: 'molding', 6: 'deco', 7: 'column', 8: 'arch', 9: 'drainpipe', 10: 'stairs', 11: 'ground_surface', 12: 'terrain', 13: 'roof', 14: 'blinds', 15: 'outer_ceiling_surface', 16: 'interior', 17: 'other' } return label_to_names def get_split(self, split): return TUMFacadeSplit(self, split=split) def get_split_list(self, split): """Returns the list of data splits available. Args: split: A string identifying the dataset split that is usually one of 'training', 'test', 'validation', or 'all'. Returns: A dataset split object providing the requested subset of the data. Raises: ValueError: Indicates that the split name passed is incorrect. The split name should be one of 'training', 'test', 'validation', or 'all'. """ if split in ['train', 'training']: return self.trainFiles elif split in ['test', 'testing']: return self.testFiles elif split in ['val', 'validation']: return self.valFiles elif split in ['all']: return self.trainFiles + self.valFiles + self.testFiles else: raise ValueError("Invalid split {}".format(split)) def is_tested(self, attr): pass def save_test_result(self, results, attr): pass class TUMFacadeSplit(BaseDatasetSplit): def __init__(self, dataset, split='train'): super().__init__(dataset, split=split) log.info("Found {} pointclouds for {}".format(len(self.path_list), split)) def __len__(self): return len(self.path_list) def get_data(self, idx): pc_path = self.path_list[idx] data = o3d.t.io.read_point_cloud(pc_path).point points = data["positions"].numpy() points = np.float32(points) labels = data['classification'].numpy().astype(np.int32).reshape((-1,)) data = {'point': points, 'feat': None, 'label': labels} return data def get_attr(self, idx): pc_path = Path(self.path_list[idx]) pc_path = str(pc_path) name = pc_path.replace('.txt', '') parts = name.split("/") name = parts[-1] split = self.split attr = {'idx': idx, 'name': name, 'path': pc_path, 'split': split} return attr DATASET._register_module(TUMFacade)