From 53ef598ac1fab09035b46631bf60b57439f387a9 Mon Sep 17 00:00:00 2001
From: Christopher Rhodes <christopher.rhodes@embl.de>
Date: Fri, 13 Sep 2024 10:08:06 +0200
Subject: [PATCH] Object classification transfer is functional
---
model_server/base/roiset.py | 45 ++++++++++++++++++++++++++++++++-----
model_server/base/util.py | 38 +------------------------------
tests/base/test_roiset.py | 10 ++++++---
3 files changed, 48 insertions(+), 45 deletions(-)
diff --git a/model_server/base/roiset.py b/model_server/base/roiset.py
index f4829112..7b1ef9af 100644
--- a/model_server/base/roiset.py
+++ b/model_server/base/roiset.py
@@ -561,9 +561,39 @@ class RoiSet(object):
)
)[-1]
se[roi.Index] = oc
- self.set_classification(name, se)
+ self.set_classification(f'classify_by_{name}', se)
+ # TODO: typecheck RoiSet not recognized
+ def get_instance_classification(self, roiset_from, iou_min: float = 0.5):
+ """
+ Transfer instance classification labels from another RoiSet based on intersection over union (IOU) similarity
+ :param roiset_from: RoiSet source of classification labels, same shape as this RoiSet
+ :param iou_min: threshold IOU below which a label is not transferred
+ """
+ if self.acc_raw.shape != roiset_from.acc_raw.shape:
+ raise ShapeMismatchError(
+ f'Expecting two RoiSets of same shape: {self.acc_raw.shape} != {roiset_from.acc_raw.shape}')
+
+ columns = [f'classify_by_{c}' for c in roiset_from.classification_columns]
+
+ if len(columns) == 0:
+ raise MissingInstanceLabelsError('Expecting at least on instance classification channel but none found')
+
+ df_overlaps = filter_df_overlap_seg(
+ roiset_from.get_df(),
+ self.get_df()
+ )
+ df_merge = pd.merge(
+ roiset_from.get_df()[columns],
+ df_overlaps.loc[df_overlaps.seg_iou > iou_min, ['overlaps_with']],
+ left_index=True,
+ right_index=True,
+ how='inner',
+ ).set_index('overlaps_with')
+ for col in columns:
+ self.set_classification(col, df_merge[col])
+
def get_object_class_map(self, name: str) -> InMemoryDataAccessor:
"""
For a given classification result, return a map where object IDs are replaced by each object's class
@@ -781,14 +811,13 @@ class RoiSet(object):
pr = 'classify_by_'
return [c.split(pr)[1] for c in self._df.columns if c.startswith(pr)]
- def set_classification(self, cname: str, se: pd.Series):
+ def set_classification(self, colname: str, se: pd.Series):
"""
Set instance classification result as a column addition on dataframe
- :param cname: name of classification result
+ :param colname: name of classification result
:param se: series containing class information
"""
- col = f'classify_by_{cname}'
- self._df[col] = se
+ self._df[colname] = se
def run_exports(self, where: Path, channel, prefix, params: RoiSetExportParams) -> dict:
"""
@@ -952,4 +981,10 @@ class MissingSegmentationError(Error):
pass
class PatchMaskShapeError(Error):
+ pass
+
+class ShapeMismatchError(Error):
+ pass
+
+class MissingInstanceLabelsError(Error):
pass
\ No newline at end of file
diff --git a/model_server/base/util.py b/model_server/base/util.py
index 58375761..81736d48 100644
--- a/model_server/base/util.py
+++ b/model_server/base/util.py
@@ -164,40 +164,4 @@ def loop_workflow(
)
if len(failures) > 0:
- pd.DataFrame(failures).to_csv(Path(output_folder_path) / 'failures.csv')
-
-def transfer_classification(
- r1: RoiSet,
- r2: RoiSet,
- iou_min: float = 0.5,
-):
-
- if r1.acc_raw.shape != r2.acc_raw.shape:
- raise RoiSetShapeMismatchError(f'Expecting two RoiSets of same shape: {r1.acc_raw.shape} != {r2.acc_raw.shape}')
-
- classes = r1.classification_columns
- class_columns = [f'classify_by_{c}' for c in r1.classification_columns]
-
- df_overlaps = filter_df_overlap_seg(r1.get_df(), r2.get_df())
- df_merge = pd.merge(
- r1.get_df()[class_columns],
- df_overlaps.loc[df_overlaps.seg_iou > iou_min, ['overlaps_with']],
- left_index=True,
- right_index=True,
- how='inner',
- # suffixes=['r1', 'r2']
- )
-
- print('hi')
- # for cl in classes:
- # se = pd.Series(dtype='Int64', index=r2.get_df().index)
- # se.loc[:] = None
- # se.loc[df_keep.]
- #
- # r1.loc[df]
-
-class Error(Exception):
- pass
-
-class RoiSetShapeMismatchError(Error):
- pass
\ No newline at end of file
+ pd.DataFrame(failures).to_csv(Path(output_folder_path) / 'failures.csv')
\ No newline at end of file
diff --git a/tests/base/test_roiset.py b/tests/base/test_roiset.py
index f36d1dc8..53c1f042 100644
--- a/tests/base/test_roiset.py
+++ b/tests/base/test_roiset.py
@@ -11,7 +11,6 @@ from model_server.base.roiset import RoiSet
from model_server.base.accessors import generate_file_accessor, InMemoryDataAccessor, write_accessor_data_to_file, PatchStack
from model_server.base.models import DummyInstanceSegmentationModel
from model_server.base.process import smooth
-from model_server.base.util import transfer_classification
import model_server.conf.testing as conf
data = conf.meta['image_files']
@@ -209,8 +208,13 @@ class TestRoiSetMonoProducts(BaseTestRoiSetMonoProducts, unittest.TestCase):
roiset1.classify_by('dummy_class', [0, 1], DummyInstanceSegmentationModel())
self.assertTrue('dummy_class' in roiset1.classification_columns)
- transfer_classification(roiset1, roiset2)
- self.assertTrue(False)
+ self.assertFalse('dummy_class' in roiset2.classification_columns)
+ roiset2.get_instance_classification(roiset1)
+ self.assertTrue('dummy_class' in roiset2.classification_columns)
+ self.assertLess(
+ roiset2.get_df().classify_by_dummy_class.count(),
+ roiset1.get_df().classify_by_dummy_class.count(),
+ )
def test_classify_by_with_derived_channel(self):
--
GitLab