batch_run_patches.py

from pathlib import Path
import re
from time import localtime, strftime

import pandas as pd

from extensions.chaeo.workflows import export_patches_from_multichannel_zstack
from model_server.accessors import InMemoryDataAccessor, write_accessor_data_to_file

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
    for ff in Path(where_output_root).iterdir():
        ma = re.match(f'batch-output-{yyyymmdd}-([\d]+)', ff.name)
        if ma:
            idx = max(idx, int(ma.groups()[0]) + 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'

    #TODO: try/catch blocks and error handling around workflow calls
    #TODO: pack JSON-serializable workflow inputs
    for ff in where_czi.iterdir():

        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),
        }

        result = export_patches_from_multichannel_zstack(**export_kwargs)

        # parse and record results
        df = result['dataframe']
        df['source_path'] = ff
        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])
            )