Working on export zstack of object labels for ilastik ingest

extensions/chaeo/examples/batch_run_patches.py

@@ -24,11 +24,10 @@ if __name__ == '__main__':
         'zmask_expand_box_by': (128, 3),
         'export_pixel_probabilities': False,
         'export_2d_patches_for_training': True,
-        'export_2d_patches_for_annotation': False,
+        'export_2d_patches_for_annotation': True,
         'export_3d_patches': False,
         'export_annotated_zstack': False,
-        'export_patch_masks': False,
-        'export_patch_label_maps': True,
+        'export_patch_masks': True,
     }
 
     input_files = get_matching_files(where_czi, 'czi', coord_filter={'P': (0, 10)}, )

extensions/chaeo/examples/label_patches.py

+from pathlib import Path
+
+from extensions.chaeo.util import autonumber_new_directory, get_matching_files, loop_workflow
+from extensions.chaeo.workflows import transfer_ecotaxa_labels_to_patch_stacks
+
+if __name__ == '__main__':
+    root = Path('c:/Users/rhodes/projects/proj0011-plankton-seg/exp0009/output')
+    transfer_ecotaxa_labels_to_patch_stacks(
+        where_masks=(root / 'batch-output-20231012-0016/patch_masks').__str__(),
+        object_csv=(root / 'batch-output-20231011-0003/df_objects.csv').__str__(),
+        ecotaxa_tsv='c:/Users/rhodes/projects/proj0011-plankton-seg/exp0013/ecotaxa_export_10468_20231012_0930.tsv',
+        where_output=autonumber_new_directory(root, 'labeled_patches')
+    )

extensions/chaeo/util.py

@@ -14,7 +14,7 @@ def autonumber_new_directory(where: str, prefix: str) -> str:
         ma = re.match(f'{prefix}-{yyyymmdd}-([\d]+)', ff.name)
         if ma:
             idx = max(idx, int(ma.groups()[0]) + 1)
-    new_path = (Path(where) / f'batch-output-{yyyymmdd}-{idx:04d}')
+    new_path = (Path(where) / f'{prefix}-{yyyymmdd}-{idx:04d}')
     new_path.mkdir(parents=True, exist_ok=False)
     return new_path.__str__()
 

extensions/chaeo/workflows.py

@@ -2,6 +2,7 @@ from pathlib import Path
 from typing import Dict
 from uuid import uuid4
 
+import numpy as np
 import pandas as pd
 
 from extensions.ilastik.models import IlastikPixelClassifierModel
@@ -30,7 +31,6 @@ def export_patches_from_multichannel_zstack(
         export_3d_patches=True,
         export_annotated_zstack=True,
         export_patch_masks=True,
-        export_patch_label_maps=True,
 ) -> Dict:
     ti = Timer()
     stack = generate_file_accessor(Path(input_zstack_path))
@@ -94,9 +94,11 @@ def export_patches_from_multichannel_zstack(
             focus_metric='max_sobel',
             ch_white=4,
             ch_rgb_overlay=(3, None, None),
-            draw_bounding_box=True,
+            draw_bounding_box=False,
             bounding_box_channel=1,
             bounding_box_linewidth=2,
+            # draw_contour=True,
+            draw_mask=True,
             overlay_gain=(0.1, 1.0, 1.0)
         )
         df_patches = pd.DataFrame(files)
@@ -123,10 +125,13 @@ def export_patches_from_multichannel_zstack(
         # prepopulate patch UUID
         df['patch_id'] = df.apply(lambda _: uuid4(), axis=1)
 
+
     if export_patch_masks:
         files = export_patch_masks_from_zstack(
             Path(where_output) / 'patch_masks',
+            stack.get_one_channel_data(4),
             zmask_meta,
+            prefix=fstem,
         )
 
     if export_annotated_zstack:
@@ -162,22 +167,61 @@ def export_patches_from_multichannel_zstack(
         'interm': interm,
     }
 
-def transfer_ecotaxa_labels_to_patch_object_maps(
-        path_to_patches: str,
-        path_to_ecotaxa_tsv: str,
-        path_output: str,
+def transfer_ecotaxa_labels_to_patch_stacks(
+        where_masks: str,
+        object_csv: str,
+        ecotaxa_tsv: str,
+        where_output: str,
+        patch_size: tuple = (256, 256),
 ) -> Dict:
-    where_patches = Path(path_to_patches)
-    df_meta = pd.read_csv(
-        path_to_ecotaxa_tsv,
+    df_obj = pd.read_csv(
+        object_csv,
+    )
+    df_ecotaxa = pd.read_csv(
+        ecotaxa_tsv,
         sep='\t',
-        header=[0, 1],
+        header=[0],
         dtype={
             ('object_annotation_date', '[t]'): str,
             ('object_annotation_time', '[t]'): str,
             ('object_annotation_category_id', '[t]'): str,
         }
     )
-    for pp in where_patches.iterdir():
-        patch = generate_file_accessor(pp)
+    df_merge = pd.merge(df_obj, df_ecotaxa, left_on='patch_id', right_on='object_id')
+    se_unique = pd.Series(
+        df_merge.object_annotation_hierarchy.unique()
+    )
+    df_split = (
+        se_unique.str.rsplit(
+            pat='>', n=1, expand=True
+        )
+    )
+    df_labels = pd.DataFrame({
+        'annotation_class_id': df_split.index,
+        'hierarchy': se_unique,
+        'annotation_class': df_split.loc[:, 1].str.lower()
+    })
+    df_labels.to_csv(Path(where_output) / 'labels_key.csv')
+
+    df_pf = pd.merge(
+        df_merge[['patch_filename', 'object_annotation_hierarchy']],
+        df_labels,
+        left_on='object_annotation_hierarchy',
+        right_on='hierarchy',
+    )
+    df_pl = df_pf[df_pf['object_annotation_hierarchy'].notnull()]
+
+    zstack = np.zeros((*patch_size, 1, len(df_pl)), dtype='uint16')
+
+    # export patches as z-stack
+    for fi, pl in enumerate(df_pl.itertuples(name='PatchFile')):
+        fn = pl._asdict()['patch_filename']
+        ac = pl._asdict()['annotation_class_id']
+        bm = generate_file_accessor(Path(where_masks) / fn).data
+        assert bm.shape == patch_size, f'Unexpected patch size {patch_size}'
+        zstack[:, :, 0, fi] = (bm == 255) * ac
+
+    # export masks as z-stack
+    write_accessor_data_to_file(where_output / 'zstack_object_label.tif', InMemoryDataAccessor(zstack))
+