Removed example, workflow, and product-generating method that exported

model_server/extensions/chaeo/examples/export_patch_focus_metrics.py

-from pathlib import Path
-import re
-from time import localtime, strftime
-from typing import Dict
-
-import pandas as pd
-
-from model_server.extensions.ilastik.models import IlastikPixelClassifierModel
-from model_server.extensions.chaeo.products import export_3d_patches_with_focus_metrics, export_patches_from_zstack
-from model_server.extensions.chaeo.zmask import RoiSet
-from model_server.base.accessors import generate_file_accessor, InMemoryDataAccessor, write_accessor_data_to_file
-from model_server.base.workflows import Timer
-
-
-def export_patch_focus_metrics_from_multichannel_zstack(
-        input_zstack_path: str,
-        ilastik_project_file: str,
-        pxmap_threshold: float,
-        pixel_class: int,
-        zmask_channel: int,
-        patches_channel: int,
-        where_output: str,
-        mask_type: str = 'boxes',
-        zmask_filters: Dict = None,
-        zmask_expand_box_by: int = None,
-        annotate_focus_metric=None,
-        **kwargs,
-) -> Dict:
-
-    ti = Timer()
-    stack = generate_file_accessor(Path(input_zstack_path))
-    fstem = Path(input_zstack_path).stem
-    ti.click('file_input')
-    assert stack.nz > 1, 'Expecting z-stack'
-
-    # MIP and classify pixels
-    mip = InMemoryDataAccessor(
-        stack.get_one_channel_data(channel=0).data.max(axis=-1, keepdims=True)
-    )
-    px_model = IlastikPixelClassifierModel(
-        params={'project_file': Path(ilastik_project_file)}
-    )
-    pxmap, _ = px_model.infer(mip)
-    ti.click('infer_pixel_probability')
-
-    obmask = InMemoryDataAccessor(
-        pxmap.data > pxmap_threshold
-    )
-    ti.click('threshold_pixel_mask')
-
-    # make zmask
-    # zmask, zmask_meta, df, interm = build_zmask_from_object_mask(
-    #     obmask.get_one_channel_data(pixel_class),
-    #     stack.get_one_channel_data(zmask_channel),
-    #     mask_type=mask_type,
-    #     filters=zmask_filters,
-    #     expand_box_by=zmask_expand_box_by,
-    # )
-    obj_table = RoiSet(
-        obmask.get_one_channel_data(pixel_class),
-        stack.get_one_channel_data(zmask_channel),
-        mask_type=mask_type,
-        filters=zmask_filters,
-        expand_box_by=zmask_expand_box_by,
-    )
-    zmask_acc = InMemoryDataAccessor(obj_table.zmask)
-    ti.click('generate_zmasks')
-
-    files = export_3d_patches_with_focus_metrics(
-        Path(where_output) / '3d_patches',
-        stack.get_one_channel_data(patches_channel),
-        obj_table.zmask_meta,
-        prefix=fstem,
-        rescale_clip=0.0,
-        make_3d=True,
-        annotate_focus_metric=annotate_focus_metric,
-    )
-    ti.click('export_3d_patches')
-
-    files = export_patches_from_zstack(
-        Path(where_output) / '2d_patches',
-        stack.get_one_channel_data(patches_channel),
-        obj_table.zmask_meta,
-        prefix=fstem,
-        draw_bounding_box=True,
-        rescale_clip=0.0,
-        # focus_metric=lambda x: np.max(sobel(x)),
-        focus_metric='max_sobel',
-        make_3d=False,
-    )
-    ti.click('export_2d_patches')
-
-    return {
-        'pixel_model_id': px_model.model_id,
-        'input_filepath': input_zstack_path,
-        'number_of_objects': len(obj_table.zmask_meta),
-        'success': True,
-        'timer_results': ti.events,
-        'dataframe': df,
-        'interm': obj_table.interm,
-    }
-
-if __name__ == '__main__':
-    where_czi = Path(
-        'c:/Users/rhodes/projects/proj0004-marine-photoactivation/data/exp0038/AutoMic/20230906-163415/Selection'
-    )
-
-    where_output_root = Path(
-        'c:/Users/rhodes/projects/proj0011-plankton-seg/exp0009/output'
-    )
-    yyyymmdd = strftime('%Y%m%d', localtime())
-    idx = 0
-    while Path(where_output_root / f'batch-output-{yyyymmdd}-{idx:04d}').exists():
-        idx += 1
-    where_output = Path(
-        where_output_root / f'batch-output-{yyyymmdd}-{idx:04d}'
-    )
-
-    csv_args = {'mode': 'w', 'header': True} # when creating file
-    px_ilp = Path.home() / 'model_server' / 'ilastik' / 'AF405-bodies_boundaries.ilp'
-
-    for ff in where_czi.iterdir():
-        if ff.stem != 'Selection--W0000--P0009-T0001':
-            continue
-
-        pattern = 'Selection--W([\d]+)--P([\d]+)-T([\d]+)'
-        ma = re.match(pattern, ff.stem)
-
-        print(ff)
-        if not ff.suffix.upper() == '.CZI':
-            continue
-        if int(ma.groups()[1]) > 10: # skip second half of set
-            continue
-
-        export_kwargs = {
-            'input_zstack_path': (where_czi / ff).__str__(),
-            'ilastik_project_file': px_ilp.__str__(),
-            'pxmap_threshold': 0.25,
-            'pixel_class': 0,
-            'zmask_channel': 0,
-            'patches_channel': 4,
-            'where_output': where_output.__str__(),
-            'mask_type': 'boxes',
-            'zmask_filters': {'area': (1e3, 1e8)},
-            'zmask_expand_box_by': (128, 3),
-            'annotate_focus_metric': 'max_sobel'
-        }
-
-        result = export_patch_focus_metrics_from_multichannel_zstack(**export_kwargs)
-
-        # parse and record results
-        df = result['dataframe']
-        df['filename'] = ff.name
-        df.to_csv(where_output / 'df_objects.csv', **csv_args)
-        pd.DataFrame(result['timer_results'], index=[0]).to_csv(where_output / 'timer_results.csv', **csv_args)
-        pd.json_normalize(export_kwargs).to_csv(where_output / 'workflow_params.csv', **csv_args)
-        csv_args = {'mode': 'a', 'header': False} # append to CSV from here on
-
-        # export intermediate data if flagged
-        for k in result['interm'].keys():
-            write_accessor_data_to_file(
-                where_output / k / (ff.stem + '.tif'),
-                InMemoryDataAccessor(result['interm'][k])
-            )
\ No newline at end of file

model_server/extensions/chaeo/products.py

@@ -97,7 +97,7 @@ def get_patches_from_zmask_meta(
     patches = []
 
     # for mi in zmask_meta:
-    for i, roi in enumerate(roiset.get_df().itertuples()):
+    for i, roi in enumerate(roiset.get_df().itertuples()): # TODO: call RoiSet.iter() when implemented
 
         # sl = roi['slice']
         # rbb = mi['relative_bounding_box'] # TODO: call rel_ fields in DF
@@ -204,7 +204,7 @@ def export_patches_from_zstack(
     )
 
     exported = []
-    for i, roi in enumerate(roiset.get_df().itertuples()):
+    for i, roi in enumerate(roiset.get_df().itertuples()):  # just used for label info
     # for i in range(0, len(zmask_meta)):
     #     mi = zmask_meta[i]
         patch = patches_acc.iat_yxcz(i)
@@ -224,86 +224,6 @@ def export_patches_from_zstack(
         })
     return exported
 
-def export_3d_patches_with_focus_metrics(
-        where: Path,
-        stack: GenericImageDataAccessor,
-        zmask_meta: list,
-        rescale_clip: float = 0.0,
-        pad_to: int = 256,
-        prefix='patch',
-        **kwargs
-):
-    """
-    Export 3D patches as multi-level z-stacks, along with CSV of various focus methods for each z-position
-
-    :param kwargs:
-    annotate_focus_metric: name focus metric to use when drawing bounding box at optimal focus z-position
-    :return:
-        list of exported files
-    """
-    assert stack.chroma == 1, 'Expecting monochromatic image data'
-    assert stack.nz > 1, 'Expecting z-stack'
-
-    def get_zstack_focus_metrics(zs):
-        nz = zs.shape[3]
-        me = _focus_metrics()
-        dd = {}
-        for zi in range(0, nz):
-            spf = zs[:, :, :, zi]
-            dd[zi] = {k: me[k](spf) for k in me.keys()}
-        return dd
-
-    exported = []
-    patch_meta = []
-    for mi in zmask_meta:
-        obj = mi['info']
-        sl = mi['slice']
-        rbb = mi['relative_bounding_box'] # TODO: use rel_ fields in DF
-        idx = mi['df_index']
-
-        patch = stack.data[sl]
-
-        assert len(patch.shape) == 4
-        assert patch.shape[2] == stack.chroma
-
-        if rescale_clip is not None:
-            patch = rescale(patch, rescale_clip)
-
-        # unpack relative bounding box and define subset of patch data
-        x0 = rbb['x0']
-        y0 = rbb['y0']
-        x1 = rbb['x1']
-        y1 = rbb['y1']
-        sp_sl = np.s_[y0: y1, x0: x1, :, :]
-        subpatch = patch[sp_sl]
-
-        # compute focus metrics for all z-levels
-        me_dict = get_zstack_focus_metrics(subpatch)
-        patch_meta.append({'label': obj.label, 'zi': obj.zi, 'metrics': me_dict})
-        me_df = pd.DataFrame(me_dict).T
-
-        # drawing bounding box only on focused slice
-        ak = kwargs.get('annotate_focus_metric')
-        if ak and ak in me_df.columns:
-            zi_foc = me_df.idxmax().to_dict()[ak]
-            patch[:, :, 0, zi_foc] = draw_box_on_patch(
-                patch[:, :, 0, zi_foc],
-                ((x0, y0), (x1, y1)),
-            )
-
-        if pad_to:
-            patch = pad(patch, pad_to)
-
-        fstem = f'{prefix}-la{obj.label:04d}-zi{obj.zi:04d}'
-        write_patch_to_file(where, fstem + '.tif', resample_to_8bit(patch))
-        me_df.to_csv(where / (fstem + '.csv'))
-        exported.append({
-            'df_index': idx,
-            'patch_filename': fstem + '.tif',
-            'focus_metrics_filename': fstem + '.csv',
-        })
-
-    return exported
 
 def export_multichannel_patches_from_zstack(
     where: Path,