↓ 3 callersFunction_focal_dice_lossSAM's mask supervision: focal + dice, ≈20:1 (focal:dice). ``logits`` and ``target`` are ``(1, 1, H, W)`` on the same device; target is float
src/digitalsreeni_image_annotator/training/sam_trainer.py:127
↓ 3 callersFunctionexport_pascal_voc_bbox(all_annotations, class_mapping, image_paths, slices, image_slices, output_dir)
src/digitalsreeni_image_annotator/io/export_formats.py:659
↓ 2 callersMethod__init__(self, input_files, output_dir, patch_size, overlap, dimensions)
src/digitalsreeni_image_annotator/dialogs/image_patcher.py:55