fruit_project.utils.datasets.alb_mosaic_dataset¶
Classes¶
Dataset wrapper that applies Albumentations' native Mosaic augmentation, |
Functions¶
Module Contents¶
- class fruit_project.utils.datasets.alb_mosaic_dataset.AlbumentationsMosaicDataset(dataset: fruit_project.utils.datasets.det_dataset.DET_DS, current_epoch: int = 0, hard_transforms: albumentations.Compose = None, easy_transforms: albumentations.Compose = None, cfg=None)[source]¶
Bases:
torch.utils.data.Dataset
Dataset wrapper that applies Albumentations’ native Mosaic augmentation, following the correct API based on official documentation.
- should_apply_mosaic() bool [source]¶
Determine if mosaic should be applied based on epoch and probability.
- _validate_and_clip_bbox(bbox: List[float], img_width: int, img_height: int) List[float] | None [source]¶
Validate and clip bounding box coordinates.
- _prepare_mosaic_metadata(primary_idx: int) List[Dict[str, Any]] [source]¶
Prepare metadata for Albumentations Mosaic transform. This now returns a LIST OF DICTIONARIES, as required by the docs.
- _apply_mosaic_augmentation(idx: int) Tuple[numpy.ndarray, List, List] [source]¶
Apply Albumentations Mosaic transform.
- fruit_project.utils.datasets.alb_mosaic_dataset.create_albumentations_mosaic_dataset(dataset: fruit_project.utils.datasets.det_dataset.DET_DS, hard_transforms: albumentations.Compose = None, easy_transforms: albumentations.Compose = None, cfg=None) AlbumentationsMosaicDataset [source]¶