search

FAU-DLM / wsi_processing_pipeline

This library helps with the key pre- and postprocessing steps necessary to use whole-slide images in deep-learning/ai projects.
4 stars 3 forks source link

Transforms not applied when using custom TileImageBlock #1

Closed ChristophNeuner closed 4 years ago

ChristophNeuner commented 4 years ago

Example:

data = fastai2.data.block.DataBlock(
        blocks=(TileImageBlock, fastai2.data.block.MultiCategoryBlock),
        get_x=lambda x: x, 
        get_y=lambda x: x.classification_labels,
        splitter=fastai2.data.transforms.FuncSplitter(lambda x: x.is_valid),
        item_tfms = fastai2.vision.augment.Resize(sz),
        batch_tfms=
[fastai2.data.transforms.Normalize.from_stats(*fastai2.vision.all.imagenet_stats)].append(fastai2.vision.augment.aug_transforms(flip_vert=True)))

dls = data.dataloaders(object_manager.get_all_top_tiles(), 
                       bs=bs)

When images are loaded with these dataloaders, the transforms are not applied.