graph LR
Indoor3DSemSegLoader["Indoor3DSemSegLoader"]
ModelNet40Loader["ModelNet40Loader"]
data_utils["data_utils"]
_build_dataloader["_build_dataloader"]
train_dataloader["train_dataloader"]
val_dataloader["val_dataloader"]
Indoor3DSemSegLoader -- "provides data to" --> _build_dataloader
ModelNet40Loader -- "provides data to" --> _build_dataloader
ModelNet40Loader -- "utilizes" --> data_utils
train_dataloader -- "calls" --> _build_dataloader
val_dataloader -- "calls" --> _build_dataloader
This subsystem is primarily responsible for efficient data loading and preprocessing for 3D point cloud datasets, specifically for the Indoor3D Semantic Segmentation and ModelNet40 datasets. The Indoor3DSemSegLoader and ModelNet40Loader components act as specialized data providers, each handling the unique aspects of their respective datasets. The data_utils component offers reusable data augmentation and normalization functionalities, which are leveraged by the dataset loaders. The central _build_dataloader function standardizes the creation of PyTorch DataLoader instances, ensuring consistent batching and shuffling. Finally, train_dataloader and val_dataloader serve as orchestrators, initiating the data pipelines for training and validation phases by utilizing _build_dataloader. This modular design ensures clear separation of concerns, allowing for easy integration of new datasets and flexible data processing workflows.
Specializes in loading and initial processing of the Indoor3D Semantic Segmentation dataset. It handles dataset-specific file parsing and data retrieval, acting as a concrete implementation of a dataset interface.
Related Classes/Methods:
Specializes in loading and initial preprocessing of the ModelNet40 dataset. It includes methods for setting the number of points and performing point cloud normalization (pc_normalize), serving as another concrete dataset loader.
Related Classes/Methods:
A utility module providing generic functions for common data augmentation (e.g., _get_angles, angle_axis) and normalization operations. These functions are designed to be reusable across different datasets and loaders.
Related Classes/Methods:
A core helper function that encapsulates the common logic for constructing PyTorch DataLoader instances. It takes processed datasets (from Indoor3DSemSegLoader or ModelNet40Loader) and configures batching, shuffling, and other DataLoader parameters, centralizing data batching logic.
Related Classes/Methods:
Orchestrates the creation of the training data pipeline. It configures and invokes _build_dataloader to produce the DataLoader for the training phase, potentially integrating data augmentation steps. This component acts as the entry point for training data access.
Related Classes/Methods:
Orchestrates the creation of the validation data pipeline. Similar to train_dataloader, it configures and invokes _build_dataloader for the validation phase. This component acts as the entry point for validation data access.
Related Classes/Methods: