deepvisiontools 0.0.4__tar.gz → 0.1.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/PKG-INFO +17 -13
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/README.md +1 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/pyproject.toml +13 -10
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/__init__.py +3 -1
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/dataset.py +5 -3
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/formats/base_data.py +7 -1
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/formats/base_formats.py +7 -2
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/inference/__init__.py +2 -1
- deepvisiontools-0.1.0/src/deepvisiontools/inference/evaluator.py +256 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/inference/predictor.py +39 -8
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/metrics/available_metrics.py +1 -1
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/metrics/base_metric.py +229 -17
- deepvisiontools-0.1.0/src/deepvisiontools/metrics/func.py +36 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools.egg-info/PKG-INFO +17 -13
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools.egg-info/SOURCES.txt +1 -0
- deepvisiontools-0.1.0/src/deepvisiontools.egg-info/requires.txt +10 -0
- deepvisiontools-0.0.4/src/deepvisiontools/metrics/func.py +0 -34
- deepvisiontools-0.0.4/src/deepvisiontools.egg-info/requires.txt +0 -9
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/LICENSE.txt +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/setup.cfg +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/setup.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/config/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/config/config.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/config/errors.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/additional_augmentations.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/augmentation_class.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/batch_augmentations.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/data_reader/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/data_reader/readers.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/errors.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/formats/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/formats/errors.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/formats/formats.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/formats/utils.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/inference/patchifier.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/metrics/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/metrics/matcher.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/metrics/utils.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/basemodel.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/mask2former/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/mask2former/mask2former.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/smp/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/smp/errors.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/smp/smp.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8/errors.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8/utils.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8/yolov8.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8seg/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8seg/errors.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8seg/utils.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8seg/yolov8seg.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/preprocessing/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/preprocessing/image.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/preprocessing/preprocessing.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/train/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/train/trainer.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/utils/__init__.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/utils/errors.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/utils/visualization.py +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools.egg-info/dependency_links.txt +0 -0
- {deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools.egg-info/top_level.txt +0 -0
|
@@ -1,8 +1,8 @@
|
|
|
1
|
-
Metadata-Version: 2.
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
2
|
Name: deepvisiontools
|
|
3
|
-
Version: 0.0
|
|
4
|
-
Summary: Overlay of PyTorch to generalize trainning and inference processes for detection
|
|
5
|
-
Author-email: Lucas Bernigaud Samatan <lucas.bernigaud-samatan@inrae.fr>, Jordan Bernigaud Samatan <jordan.bernigaud-samatan@inrae.fr>
|
|
3
|
+
Version: 0.1.0
|
|
4
|
+
Summary: Overlay of PyTorch to generalize trainning and inference processes for detection, instance segmentation and semantic segmentation tasks.
|
|
5
|
+
Author-email: Lucas Bernigaud Samatan <lucas.bernigaud-samatan@inrae.fr>, Jordan Bernigaud Samatan <jordan.bernigaud-samatan@inrae.fr>, Paul Faucher <paul.faucher@inrae.fr>
|
|
6
6
|
Maintainer-email: Jordan Bernigaud Samatan <jordan.bernigaud-samatan@inrae.fr>
|
|
7
7
|
License: GNU AFFERO GENERAL PUBLIC LICENSE
|
|
8
8
|
Version 3, 19 November 2007
|
|
@@ -670,20 +670,23 @@ Project-URL: Homepage, https://forgemia.inra.fr/ue-apc/librairies/python/deepvis
|
|
|
670
670
|
Project-URL: Issues, https://forgemia.inra.fr/ue-apc/librairies/python/deepvisiontools_v1/-/issues
|
|
671
671
|
Project-URL: Documentation, https://deepvisiontools.readthedocs.io/en/latest/
|
|
672
672
|
Project-URL: Repository, https://forgemia.inra.fr/ue-apc/librairies/python/deepvisiontools_v1
|
|
673
|
-
Classifier: Development Status ::
|
|
673
|
+
Classifier: Development Status :: 4 - Beta
|
|
674
674
|
Classifier: Programming Language :: Python :: 3.11
|
|
675
|
-
|
|
675
|
+
Classifier: Programming Language :: Python :: 3.12
|
|
676
|
+
Requires-Python: <=3.12.9,>=3.11.9
|
|
676
677
|
Description-Content-Type: text/markdown
|
|
677
678
|
License-File: LICENSE.txt
|
|
678
|
-
Requires-Dist: torch
|
|
679
|
-
Requires-Dist: torchvision
|
|
679
|
+
Requires-Dist: torch<=2.6.0,>=2.4.0
|
|
680
|
+
Requires-Dist: torchvision<=0.21.0,>=0.19.0
|
|
680
681
|
Requires-Dist: pycocotools==2.0.8
|
|
681
|
-
Requires-Dist: torchmetrics
|
|
682
|
-
Requires-Dist: tensorboard
|
|
683
|
-
Requires-Dist: ultralytics
|
|
684
|
-
Requires-Dist: transformers
|
|
682
|
+
Requires-Dist: torchmetrics<=1.7.0,>=1.4.0
|
|
683
|
+
Requires-Dist: tensorboard<=2.19.0,>=2.18.0
|
|
684
|
+
Requires-Dist: ultralytics<=8.3.95,>=8.3.28
|
|
685
|
+
Requires-Dist: transformers<=4.50.0,>=4.44.2
|
|
685
686
|
Requires-Dist: colorcet==3.1.0
|
|
686
|
-
Requires-Dist: segmentation-models-pytorch
|
|
687
|
+
Requires-Dist: segmentation-models-pytorch<=0.4.0,>=0.3.4
|
|
688
|
+
Requires-Dist: openpyxl==3.1.5
|
|
689
|
+
Dynamic: license-file
|
|
687
690
|
|
|
688
691
|
<!--
|
|
689
692
|
<p align="center">
|
|
@@ -713,6 +716,7 @@ Similar stands for other models.
|
|
|
713
716
|
All documentation about deepvisiontools, including tutorials, can be found at : https://deepvisiontools.readthedocs.io/en/latest/
|
|
714
717
|
|
|
715
718
|
## Installation
|
|
719
|
+
We have tested deepvisiontools for python versions between 3.11.9 and 3.12.9
|
|
716
720
|
|
|
717
721
|
```shell
|
|
718
722
|
pip install deepvisiontools
|
|
@@ -26,6 +26,7 @@ Similar stands for other models.
|
|
|
26
26
|
All documentation about deepvisiontools, including tutorials, can be found at : https://deepvisiontools.readthedocs.io/en/latest/
|
|
27
27
|
|
|
28
28
|
## Installation
|
|
29
|
+
We have tested deepvisiontools for python versions between 3.11.9 and 3.12.9
|
|
29
30
|
|
|
30
31
|
```shell
|
|
31
32
|
pip install deepvisiontools
|
|
@@ -7,28 +7,31 @@ name = "deepvisiontools"
|
|
|
7
7
|
authors = [
|
|
8
8
|
{ name = "Lucas Bernigaud Samatan", email = "lucas.bernigaud-samatan@inrae.fr" },
|
|
9
9
|
{ name = "Jordan Bernigaud Samatan", email = "jordan.bernigaud-samatan@inrae.fr" },
|
|
10
|
+
{ name = "Paul Faucher", email = "paul.faucher@inrae.fr" },
|
|
10
11
|
]
|
|
11
12
|
maintainers = [
|
|
12
13
|
{ name = "Jordan Bernigaud Samatan", email = "jordan.bernigaud-samatan@inrae.fr" },
|
|
13
14
|
]
|
|
14
|
-
description = "Overlay of PyTorch to generalize trainning and inference processes for detection
|
|
15
|
+
description = "Overlay of PyTorch to generalize trainning and inference processes for detection, instance segmentation and semantic segmentation tasks."
|
|
15
16
|
readme = "README.md"
|
|
16
17
|
classifiers = [
|
|
17
|
-
"Development Status ::
|
|
18
|
+
"Development Status :: 4 - Beta",
|
|
18
19
|
"Programming Language :: Python :: 3.11",
|
|
20
|
+
"Programming Language :: Python :: 3.12"
|
|
19
21
|
]
|
|
20
22
|
dynamic = ["version"]
|
|
21
|
-
requires-python = ">=3.11.9"
|
|
23
|
+
requires-python = ">=3.11.9,<=3.12.9"
|
|
22
24
|
dependencies = [
|
|
23
|
-
"torch
|
|
24
|
-
"torchvision
|
|
25
|
+
"torch>=2.4.0,<=2.6.0",
|
|
26
|
+
"torchvision>=0.19.0,<=0.21.0",
|
|
25
27
|
"pycocotools==2.0.8",
|
|
26
|
-
"torchmetrics
|
|
27
|
-
"tensorboard
|
|
28
|
-
"ultralytics
|
|
29
|
-
"transformers
|
|
28
|
+
"torchmetrics>=1.4.0,<=1.7.0",
|
|
29
|
+
"tensorboard>=2.18.0,<=2.19.0",
|
|
30
|
+
"ultralytics>=8.3.28,<=8.3.95",
|
|
31
|
+
"transformers>=4.44.2,<=4.50.0",
|
|
30
32
|
"colorcet==3.1.0",
|
|
31
|
-
"segmentation-models-pytorch
|
|
33
|
+
"segmentation-models-pytorch>=0.3.4,<=0.4.0",
|
|
34
|
+
"openpyxl==3.1.5"
|
|
32
35
|
]
|
|
33
36
|
|
|
34
37
|
license = { file = "LICENSE.txt" }
|
|
@@ -3,8 +3,9 @@ from deepvisiontools.train.trainer import Trainer
|
|
|
3
3
|
from deepvisiontools.data.dataset import DeepVisionDataset, DeepVisionLoader
|
|
4
4
|
from deepvisiontools.utils import visualization
|
|
5
5
|
from deepvisiontools.inference.predictor import Predictor
|
|
6
|
+
from deepvisiontools.inference.evaluator import Evaluator
|
|
6
7
|
|
|
7
|
-
__version__ = "0.0
|
|
8
|
+
__version__ = "0.1.0"
|
|
8
9
|
|
|
9
10
|
__all__ = (
|
|
10
11
|
Configuration,
|
|
@@ -13,4 +14,5 @@ __all__ = (
|
|
|
13
14
|
DeepVisionLoader,
|
|
14
15
|
visualization,
|
|
15
16
|
Predictor,
|
|
17
|
+
Evaluator,
|
|
16
18
|
)
|
|
@@ -191,19 +191,21 @@ class DeepVisionDataset(Dataset):
|
|
|
191
191
|
dataset3._indexes = idx[stop2:]
|
|
192
192
|
return dataset1, dataset2, dataset3
|
|
193
193
|
|
|
194
|
-
def keep_indexes(self, indexes: Union[list, slice, Tensor]):
|
|
194
|
+
def keep_indexes(self, indexes: Union[list, slice, Tensor]) -> DeepVisionDataset:
|
|
195
195
|
"""Filter dataset by keeping only indices given in arg.
|
|
196
196
|
|
|
197
197
|
Args:
|
|
198
198
|
indexes (``Union[list, slice, Tensor]``): can be slice, Tensor or list. To use slice please use : slice(i, j) with i, j desired slice indexes in arg.
|
|
199
199
|
"""
|
|
200
|
+
dataset = copy.deepcopy(self)
|
|
200
201
|
if isinstance(indexes, Tensor):
|
|
201
202
|
assert (
|
|
202
203
|
indexes.dim() == 1
|
|
203
204
|
), f"Must use Tensor of dim 1 for indexes, got {indexes.shape}"
|
|
204
205
|
indexes = indexes.tolist()
|
|
205
|
-
indx = torch.tensor(
|
|
206
|
-
|
|
206
|
+
indx = torch.tensor(dataset._indexes)[indexes].tolist()
|
|
207
|
+
dataset._indexes = indx
|
|
208
|
+
return dataset
|
|
207
209
|
|
|
208
210
|
def export_dataset(
|
|
209
211
|
self,
|
|
@@ -144,7 +144,13 @@ class SemanticMaskData(BaseData):
|
|
|
144
144
|
mask = Mask(mask[0])
|
|
145
145
|
self._value = copy.deepcopy(mask) # convert to long
|
|
146
146
|
self._canvas_size = (mask.shape[-2], mask.shape[-1])
|
|
147
|
-
|
|
147
|
+
# compute nb of objects (unique cls ids) in mask that are not background (i.e 0)
|
|
148
|
+
nb_object = (
|
|
149
|
+
torch.unique(mask).shape[0] - 1
|
|
150
|
+
if 0 in torch.unique(mask).tolist()
|
|
151
|
+
else torch.unique(mask).shape[0]
|
|
152
|
+
)
|
|
153
|
+
self._nb_object = nb_object
|
|
148
154
|
self._device = Configuration().device
|
|
149
155
|
# we use here a trick : handling scores operation is complicated at format level.
|
|
150
156
|
# We dynamically will modify it in the operationhandler class and implement the modification here
|
|
@@ -190,8 +190,13 @@ class BaseSemanticFormat(BaseFormat):
|
|
|
190
190
|
scores.shape[-2:] == data.value.shape[-2:]
|
|
191
191
|
), f"scores do not have the same shape as data.value, got {scores.shape} and {data.value.shape}"
|
|
192
192
|
self._data = data
|
|
193
|
-
# labels are derived from semantic mask data value
|
|
194
|
-
|
|
193
|
+
# labels are derived from semantic mask data value (careful : need to handle the case where 0 is not present in mask ...)
|
|
194
|
+
labs = (
|
|
195
|
+
data.value.unique()[1:].long()
|
|
196
|
+
if 0 in data.value.unique().tolist()
|
|
197
|
+
else data.value.unique().long()
|
|
198
|
+
)
|
|
199
|
+
self._labels: Tensor = labs
|
|
195
200
|
# Check score
|
|
196
201
|
if scores != None:
|
|
197
202
|
scores = (
|
|
@@ -1,4 +1,5 @@
|
|
|
1
1
|
from deepvisiontools.inference.patchifier import BasePatchifier, SemanticPatchifier, DetectPatchifier, PostProcesser
|
|
2
2
|
from deepvisiontools.inference.predictor import Predictor, PredictorDataLoader
|
|
3
|
+
from deepvisiontools.inference.evaluator import Evaluator
|
|
3
4
|
|
|
4
|
-
__all__ = ("Predictor", "BasePatchifier", "DetectPatchifier", "SemanticPatchifier", "PostProcesser", "PredictorDataLoader")
|
|
5
|
+
__all__ = ("Predictor", "BasePatchifier", "DetectPatchifier", "SemanticPatchifier", "PostProcesser", "PredictorDataLoader", "Evaluator")
|
|
@@ -0,0 +1,256 @@
|
|
|
1
|
+
from deepvisiontools import DeepVisionDataset, Configuration
|
|
2
|
+
from deepvisiontools.inference import Predictor
|
|
3
|
+
from deepvisiontools.utils import visualization
|
|
4
|
+
from tqdm import tqdm
|
|
5
|
+
from pathlib import Path
|
|
6
|
+
import pandas as pd
|
|
7
|
+
from typing import Literal, Union
|
|
8
|
+
from openpyxl import load_workbook
|
|
9
|
+
from openpyxl.styles import Font, PatternFill
|
|
10
|
+
import random
|
|
11
|
+
|
|
12
|
+
|
|
13
|
+
class Evaluator:
|
|
14
|
+
"""Evaluator class : evaluate a given Predictor (model + patch_size + additional Configurations) on a dataset with given metrics.
|
|
15
|
+
The results are saved in a generated csv file (metrics at dataset level) and in a xlsx file (metrics at dataset and sample levels).
|
|
16
|
+
Highlighting samples that deviate from the mean or median (giving deviation_method) by nb_sigma sigma in the xlsx file.
|
|
17
|
+
Create visualizations of predictions giving number_visu.
|
|
18
|
+
Returns the dictionnary with metrics at dataset level and the dictionnary with metrics at sample level.
|
|
19
|
+
Prints a dataframe with metrics at dataset level (same content as in the generated csv file).
|
|
20
|
+
|
|
21
|
+
Args:
|
|
22
|
+
predictor (``Predictor``): Predictor class to evaluate
|
|
23
|
+
metrics (``list``): List of metrics to evaluate the Predictor on
|
|
24
|
+
deviation_method (``Literal["mean", "median"]``, **optional**): method to compute outlayers. Defaults to "mean".
|
|
25
|
+
nb_sigma (``Union[int, float]``, **optional**): number of standard deviations for outlayers. Defaults to 2.
|
|
26
|
+
|
|
27
|
+
Example:
|
|
28
|
+
----------
|
|
29
|
+
|
|
30
|
+
.. highlight:: python
|
|
31
|
+
.. code-block:: python
|
|
32
|
+
|
|
33
|
+
>>> from deepvisiontools import Evaluator, Predictor
|
|
34
|
+
>>> from deepvisiontools.metrics import DetectF1Score
|
|
35
|
+
>>> predictor = Predictor(model=\path\to\model.pth)
|
|
36
|
+
>>> evaluator = Evaluator(predictor, metrics=[DetectF1Score()])
|
|
37
|
+
>>> evaluator.evaluate(mydataset, "results")
|
|
38
|
+
|
|
39
|
+
Attributes
|
|
40
|
+
----------
|
|
41
|
+
|
|
42
|
+
Attributes:
|
|
43
|
+
predictor (``Predictor``)
|
|
44
|
+
metrics (``List[BaseMetric]``)
|
|
45
|
+
nb_sigma (``int``)
|
|
46
|
+
data_type (``Literal["instance_mask", "bbox", "keypoint", "semantic_mask"]``)
|
|
47
|
+
deviation_method (```Literal["mean", "median"]``)
|
|
48
|
+
|
|
49
|
+
|
|
50
|
+
**Methods**
|
|
51
|
+
|
|
52
|
+
"""
|
|
53
|
+
|
|
54
|
+
def __init__(
|
|
55
|
+
self,
|
|
56
|
+
predictor: Predictor,
|
|
57
|
+
metrics: list,
|
|
58
|
+
deviation_method: Literal["mean", "median"] = "mean",
|
|
59
|
+
nb_sigma: Union[int, float] = 2,
|
|
60
|
+
):
|
|
61
|
+
|
|
62
|
+
self.predictor = predictor
|
|
63
|
+
self.metrics = metrics
|
|
64
|
+
self.nb_sigma = nb_sigma
|
|
65
|
+
self.data_type = Configuration().data_type
|
|
66
|
+
self.deviation_method = deviation_method
|
|
67
|
+
assert deviation_method in [
|
|
68
|
+
"mean",
|
|
69
|
+
"median",
|
|
70
|
+
], "deviation_method must be 'mean' or 'median'"
|
|
71
|
+
|
|
72
|
+
def evaluate(
|
|
73
|
+
self,
|
|
74
|
+
dataset: DeepVisionDataset,
|
|
75
|
+
result_folder: Union[str, Path],
|
|
76
|
+
number_visu: Union[Literal["all"], int] = "all",
|
|
77
|
+
):
|
|
78
|
+
"""Run evaluation on dataset. Compute metrics for dataset and for each sample of the dataset."""
|
|
79
|
+
result_folder = (
|
|
80
|
+
result_folder if isinstance(result_folder, Path) else Path(result_folder)
|
|
81
|
+
)
|
|
82
|
+
result_folder.mkdir(parents=True, exist_ok=True)
|
|
83
|
+
export_csv_path = result_folder / "results.csv"
|
|
84
|
+
export_xlsx_path = result_folder / "results.xlsx"
|
|
85
|
+
visu_folder_path = result_folder / "visu"
|
|
86
|
+
|
|
87
|
+
assert (
|
|
88
|
+
isinstance(number_visu, int) or number_visu == "all"
|
|
89
|
+
), "number_visu must be integer or equal to 'all'"
|
|
90
|
+
|
|
91
|
+
dataset.preprocessing = None # because already preprocess in Predictor
|
|
92
|
+
|
|
93
|
+
# Creation and initialisation of the dictionary containing the metrics for each image (one dictionary/metric containing one dictionary/image)
|
|
94
|
+
metric_sample_dict = {}
|
|
95
|
+
for metric in self.metrics:
|
|
96
|
+
metric.to(Configuration().device)
|
|
97
|
+
metric_sample_dict.update({metric.name: {}})
|
|
98
|
+
|
|
99
|
+
if (
|
|
100
|
+
number_visu != 0 and number_visu != "all"
|
|
101
|
+
): # randomly select index of the number_visu images to visualize
|
|
102
|
+
list_index_img = [i for i in range(len(dataset))]
|
|
103
|
+
random.shuffle(list_index_img)
|
|
104
|
+
list_index_img_visu = list_index_img[:number_visu]
|
|
105
|
+
|
|
106
|
+
# For each sample in the dataset : prediction then calculation and storage of tp, fp, tn, fn (for each metric)
|
|
107
|
+
iterator = tqdm(dataset, total=len(dataset), desc=f"Evaluation progress ")
|
|
108
|
+
|
|
109
|
+
for index_img, (image, target, img_name) in enumerate(iterator):
|
|
110
|
+
prediction = self.predictor.predict(image)
|
|
111
|
+
|
|
112
|
+
for metric in self.metrics:
|
|
113
|
+
metric.update(
|
|
114
|
+
prediction, target
|
|
115
|
+
) # update 'stats' attribute for each metric
|
|
116
|
+
|
|
117
|
+
metric_sample_dict[metric.name].update(
|
|
118
|
+
{img_name: {"img_index_in_dataset": index_img}}
|
|
119
|
+
) # creation of the image dictionary (for the given metric) with its index information
|
|
120
|
+
# compute metric for the image (all classes combined and for each class)
|
|
121
|
+
metric_sample = metric.compute_last_sample()
|
|
122
|
+
metric_sample_dict[metric.name][img_name].update(metric_sample)
|
|
123
|
+
|
|
124
|
+
# prediction visualization
|
|
125
|
+
if number_visu != 0 and number_visu != "all":
|
|
126
|
+
if index_img in list_index_img_visu:
|
|
127
|
+
visualization(
|
|
128
|
+
image,
|
|
129
|
+
prediction,
|
|
130
|
+
save_path=visu_folder_path
|
|
131
|
+
/ "visu_img{0}_{1}".format(index_img, img_name),
|
|
132
|
+
)
|
|
133
|
+
elif number_visu == "all":
|
|
134
|
+
visualization(
|
|
135
|
+
image,
|
|
136
|
+
prediction,
|
|
137
|
+
save_path=visu_folder_path
|
|
138
|
+
/ "visu_img{0}_{1}".format(index_img, img_name),
|
|
139
|
+
)
|
|
140
|
+
|
|
141
|
+
# after all updates recompute to get averaged values of metric (global, samplewise, macro, micro), save in metric_dict
|
|
142
|
+
metric_dict = {}
|
|
143
|
+
for metric in self.metrics:
|
|
144
|
+
results = metric.compute()
|
|
145
|
+
metric_dict.update({metric.name: results})
|
|
146
|
+
|
|
147
|
+
# Converts dictionary metric values from tensor to float :
|
|
148
|
+
for metric_name in metric_dict:
|
|
149
|
+
for sous_metric_name in metric_dict[metric_name]:
|
|
150
|
+
metric_dict[metric_name][sous_metric_name] = round(
|
|
151
|
+
metric_dict[metric_name][sous_metric_name].item(), 3
|
|
152
|
+
)
|
|
153
|
+
|
|
154
|
+
# Converts the dict into a dataframe, which is saved in a csv file:
|
|
155
|
+
df_metrics = pd.DataFrame(metric_dict)
|
|
156
|
+
df_metrics.to_csv(export_csv_path)
|
|
157
|
+
|
|
158
|
+
# Create the Excel file with metrics values for each image
|
|
159
|
+
sheet_names = self._create_xls(export_xlsx_path, df_metrics, metric_sample_dict)
|
|
160
|
+
|
|
161
|
+
# Modifies the Excel file with metrics : highlights extreme values and adds mean, std, median
|
|
162
|
+
self._formate_xls(export_xlsx_path, sheet_names, metric_sample_dict)
|
|
163
|
+
|
|
164
|
+
print(df_metrics)
|
|
165
|
+
return (metric_dict, metric_sample_dict)
|
|
166
|
+
|
|
167
|
+
def _create_xls(self, export_xlsx_path, df_metrics, metric_sample_dict):
|
|
168
|
+
"""Create an Excel file with metrics values for each image"""
|
|
169
|
+
sheet_names = []
|
|
170
|
+
with pd.ExcelWriter(export_xlsx_path) as writer:
|
|
171
|
+
df_metrics.to_excel(
|
|
172
|
+
writer, sheet_name="Dataset_Metrics", index=True
|
|
173
|
+
) # creates a first sheet with the 'sub-metrics' for each metric (equivalent to the contents of the csv generated earlier)
|
|
174
|
+
for metric in metric_sample_dict:
|
|
175
|
+
df_sample = pd.DataFrame(metric_sample_dict[metric]).T
|
|
176
|
+
df_sample["Outlier?"] = None
|
|
177
|
+
sheet_name = "{0}_by_samples".format(metric)
|
|
178
|
+
if len(sheet_name) > 31: # the length of excel sheets name is 31
|
|
179
|
+
sheet_name = sheet_name[: len(sheet_name) - (len(sheet_name) - 31)]
|
|
180
|
+
df_sample.to_excel(
|
|
181
|
+
writer, sheet_name=sheet_name, index=True
|
|
182
|
+
) # creates a sheet by metric with details by image
|
|
183
|
+
sheet_names.append(sheet_name)
|
|
184
|
+
return sheet_names
|
|
185
|
+
|
|
186
|
+
def _highlight_cells(self, sheet, colonne_verif, min_val, max_val):
|
|
187
|
+
"""Applies a red background and bold text if the value is outside the range"""
|
|
188
|
+
# Style for formatting (red fill + bold text)
|
|
189
|
+
red_fill = PatternFill(
|
|
190
|
+
start_color="FFC7CE", end_color="FFC7CE", fill_type="solid"
|
|
191
|
+
)
|
|
192
|
+
bold_font = Font(bold=True, color="9C0006")
|
|
193
|
+
|
|
194
|
+
for row in sheet.iter_rows(
|
|
195
|
+
min_row=2, min_col=colonne_verif, max_col=colonne_verif
|
|
196
|
+
):
|
|
197
|
+
for cell in row:
|
|
198
|
+
if cell.value is not None and (
|
|
199
|
+
cell.value < min_val or cell.value > max_val
|
|
200
|
+
):
|
|
201
|
+
cell.fill = red_fill
|
|
202
|
+
sheet.cell(row=cell.row, column=1).font = (
|
|
203
|
+
bold_font # Make the index bold and red
|
|
204
|
+
)
|
|
205
|
+
sheet.cell(cell.row, sheet.max_column).value = (
|
|
206
|
+
"True" # True in "Outlier?" column
|
|
207
|
+
)
|
|
208
|
+
|
|
209
|
+
def _formate_xls(self, export_xlsx_path, sheet_names, metric_sample_dict):
|
|
210
|
+
"""Modifies the Excel file with metrics : highlights extreme values and adds mean, std, median"""
|
|
211
|
+
|
|
212
|
+
wb = load_workbook(export_xlsx_path)
|
|
213
|
+
|
|
214
|
+
for sh_name, metric in zip(sheet_names, metric_sample_dict):
|
|
215
|
+
|
|
216
|
+
df_sample = pd.DataFrame(metric_sample_dict[metric]).T
|
|
217
|
+
columns_list = df_sample.columns.tolist()
|
|
218
|
+
sheet_maw_row = wb[sh_name].max_row
|
|
219
|
+
|
|
220
|
+
# add mean, std, median labels in last rows
|
|
221
|
+
wb[sh_name].cell(sheet_maw_row + 2, 1).value = "Mean : "
|
|
222
|
+
wb[sh_name].cell(sheet_maw_row + 3, 1).value = "Std : "
|
|
223
|
+
wb[sh_name].cell(sheet_maw_row + 4, 1).value = "Median : "
|
|
224
|
+
|
|
225
|
+
# add mean, std and median values in last rows for each class and "all_class"
|
|
226
|
+
for num_col, col_name in enumerate(columns_list[1:]):
|
|
227
|
+
mean = df_sample.mean()[col_name]
|
|
228
|
+
std = df_sample.std()[col_name]
|
|
229
|
+
median = df_sample.median()[col_name]
|
|
230
|
+
|
|
231
|
+
num_col_xsl = num_col + 3
|
|
232
|
+
# highlights extreme values
|
|
233
|
+
if self.deviation_method == "median":
|
|
234
|
+
self._highlight_cells(
|
|
235
|
+
wb[sh_name],
|
|
236
|
+
colonne_verif=num_col_xsl,
|
|
237
|
+
min_val=median - (self.nb_sigma * std),
|
|
238
|
+
max_val=median + (self.nb_sigma * std),
|
|
239
|
+
)
|
|
240
|
+
else:
|
|
241
|
+
self._highlight_cells(
|
|
242
|
+
wb[sh_name],
|
|
243
|
+
colonne_verif=num_col_xsl,
|
|
244
|
+
min_val=mean - (self.nb_sigma * std),
|
|
245
|
+
max_val=mean + (self.nb_sigma * std),
|
|
246
|
+
)
|
|
247
|
+
|
|
248
|
+
# add mean, std and median in last rows
|
|
249
|
+
wb[sh_name].cell(sheet_maw_row + 2, num_col_xsl).value = round(mean, 3)
|
|
250
|
+
wb[sh_name].cell(sheet_maw_row + 3, num_col_xsl).value = round(std, 3)
|
|
251
|
+
wb[sh_name].cell(sheet_maw_row + 4, num_col_xsl).value = round(
|
|
252
|
+
median, 3
|
|
253
|
+
)
|
|
254
|
+
|
|
255
|
+
# Save file after modifications
|
|
256
|
+
wb.save(export_xlsx_path)
|
|
@@ -16,6 +16,7 @@ from deepvisiontools.inference.patchifier import (
|
|
|
16
16
|
)
|
|
17
17
|
from torchvision.transforms.v2 import Pad, CenterCrop
|
|
18
18
|
from tqdm import tqdm
|
|
19
|
+
from deepvisiontools.formats import SemanticMaskData, SemanticMaskFormat
|
|
19
20
|
|
|
20
21
|
|
|
21
22
|
def build_patchifier(*args, **kwargs):
|
|
@@ -39,7 +40,8 @@ class Predictor:
|
|
|
39
40
|
nms_iou_threshold (``float``, **optional**): nms threshold to be used when upatchifying. Defaults to 0.45.
|
|
40
41
|
final_score_threshold (``float``, **optional**): Apply a score thresholding after penalty and after nms. Defaults to 0.4.
|
|
41
42
|
categories (``Dict[int, str]``, **optional**): To rename your categories in the visualization.
|
|
42
|
-
patchifier (``Union[BasePatchifier, None], **optional
|
|
43
|
+
patchifier (``Union[BasePatchifier, None], **optional**): If None use default SemanticPatchifier or DetectPatchifier according to Configuration().data_type. Default to None.
|
|
44
|
+
verbose (``bool``, **optional**) : if set to True will display progress state in patchs predictions. Default to True.
|
|
43
45
|
|
|
44
46
|
Example:
|
|
45
47
|
----------
|
|
@@ -64,6 +66,7 @@ class Predictor:
|
|
|
64
66
|
cropper (``Transform``)
|
|
65
67
|
patchifier (``BasePatchifier``)
|
|
66
68
|
categories (``Dict[int, str]``)
|
|
69
|
+
verbose (``bool``)
|
|
67
70
|
|
|
68
71
|
|
|
69
72
|
**Methods**
|
|
@@ -79,6 +82,7 @@ class Predictor:
|
|
|
79
82
|
batch_size: int = 1,
|
|
80
83
|
categories: Dict[int, str] = None,
|
|
81
84
|
patchifier: Union[BasePatchifier, None] = None,
|
|
85
|
+
verbose: bool = True,
|
|
82
86
|
):
|
|
83
87
|
assert any(
|
|
84
88
|
[
|
|
@@ -103,24 +107,30 @@ class Predictor:
|
|
|
103
107
|
else:
|
|
104
108
|
self.patchifier = patchifier
|
|
105
109
|
self.categories = categories
|
|
110
|
+
self.verbose = verbose
|
|
106
111
|
|
|
107
112
|
def forward_pass(self, batch_patchs: Tensor) -> BatchedFormat:
|
|
108
113
|
"""Run predictions on image / batch of patches"""
|
|
109
114
|
with torch.no_grad():
|
|
110
115
|
loader = PredictorDataLoader(batch_patchs, self.batch_size)
|
|
111
116
|
# wrap in tqdm if more than one batch
|
|
112
|
-
|
|
113
|
-
|
|
114
|
-
|
|
115
|
-
|
|
116
|
-
|
|
117
|
+
if self.verbose:
|
|
118
|
+
loader = (
|
|
119
|
+
tqdm(
|
|
120
|
+
loader, desc="Predict on batch of patches : ", total=len(loader)
|
|
121
|
+
)
|
|
122
|
+
if len(loader) > 1
|
|
123
|
+
else loader
|
|
124
|
+
)
|
|
117
125
|
predictions = BatchedFormat([])
|
|
118
126
|
for patch in loader:
|
|
119
127
|
predictions += self.model.get_predictions(patch)
|
|
120
128
|
return predictions
|
|
121
129
|
|
|
122
130
|
def predict(
|
|
123
|
-
self,
|
|
131
|
+
self,
|
|
132
|
+
image: Union[str, Path, Tensor],
|
|
133
|
+
visu_path: Union[str, Path] = "",
|
|
124
134
|
) -> BaseFormat:
|
|
125
135
|
"""Main function of ```Predictor``` : call everything needed for prediction.
|
|
126
136
|
|
|
@@ -143,12 +153,33 @@ class Predictor:
|
|
|
143
153
|
h_original, w_original = image.shape[-2:]
|
|
144
154
|
image = image.to(Configuration().device)
|
|
145
155
|
# preprocess
|
|
146
|
-
|
|
156
|
+
if self.preprocessing != None:
|
|
157
|
+
preprocessed_image = self.preprocessing(image[None, :])[0]
|
|
158
|
+
else:
|
|
159
|
+
preprocessed_image = image
|
|
147
160
|
# handles patch if needed
|
|
148
161
|
if self.patch_size != None and self.patch_size != (
|
|
149
162
|
image.shape[-2],
|
|
150
163
|
image.shape[-1],
|
|
151
164
|
):
|
|
165
|
+
_h_pad, _w_pad = 0, 0
|
|
166
|
+
if image.shape[-2] < self.patch_size[0]:
|
|
167
|
+
_h_pad = self.patch_size[0] - image.shape[-2]
|
|
168
|
+
_t_pad = int(_h_pad // 2)
|
|
169
|
+
_t_pad = _t_pad if _t_pad > 0 else 0
|
|
170
|
+
_b_pad = _h_pad - _t_pad
|
|
171
|
+
_b_pad = _b_pad if _b_pad > 0 else 0
|
|
172
|
+
if image.shape[-2] < self.patch_size[0]:
|
|
173
|
+
_w_pad = self.patch_size[1] - image.shape[-1]
|
|
174
|
+
_l_pad = int(_w_pad // 2)
|
|
175
|
+
_l_pad = _l_pad if _l_pad > 0 else 0
|
|
176
|
+
_r_pad = _w_pad - _l_pad
|
|
177
|
+
_r_pad = _r_pad if _r_pad > 0 else 0
|
|
178
|
+
if _h_pad > 0 or _w_pad > 0:
|
|
179
|
+
preprocessed_image = Pad((_l_pad, _t_pad, _r_pad, _b_pad))(
|
|
180
|
+
preprocessed_image
|
|
181
|
+
)
|
|
182
|
+
|
|
152
183
|
batch_patch, pad_origins, padded_image, pad_coord = (
|
|
153
184
|
self.patchifier.patchify(preprocessed_image)
|
|
154
185
|
)
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/metrics/available_metrics.py
RENAMED
|
@@ -118,7 +118,7 @@ class ClassifRecall(ClassifMetric):
|
|
|
118
118
|
super().__init__(func=F.recall, name="ClassifRecall", *args, **kwargs)
|
|
119
119
|
|
|
120
120
|
|
|
121
|
-
class
|
|
121
|
+
class ClassifPrecision(ClassifMetric):
|
|
122
122
|
"""Classification precision score"""
|
|
123
123
|
|
|
124
124
|
def __init__(self, *args, **kwargs):
|
|
@@ -35,12 +35,10 @@ class DetectMetric(Metric):
|
|
|
35
35
|
Configuration().data_type != "semantic_mask"
|
|
36
36
|
), "Can't use detection metrics with Configuration().data_type = semantic_mask. Must be instance_mask or bbox"
|
|
37
37
|
super().__init__(**kwargs)
|
|
38
|
-
self.add_state("tp", default=torch.tensor(0), dist_reduce_fx="sum")
|
|
39
|
-
self.add_state("fp", default=torch.tensor(0), dist_reduce_fx="sum")
|
|
40
|
-
self.add_state("fn", default=torch.tensor(0), dist_reduce_fx="sum")
|
|
41
|
-
self.add_state("samplewise", default=[], dist_reduce_fx="cat")
|
|
42
38
|
self.func = func
|
|
43
39
|
self.name = name
|
|
40
|
+
self.add_state("stats", default=[], dist_reduce_fx="cat")
|
|
41
|
+
self.nc = 1 # always one class
|
|
44
42
|
|
|
45
43
|
def update(
|
|
46
44
|
self,
|
|
@@ -73,25 +71,81 @@ class DetectMetric(Metric):
|
|
|
73
71
|
tp, fp, fn, _ = matcher.match_pred_target(pred, targ)
|
|
74
72
|
elif pred.nb_object != 0 and targ.nb_object == 0:
|
|
75
73
|
tp, fp, fn, _ = 0, pred.nb_object, 0, None
|
|
76
|
-
elif prediction.size == 0 and target.size != 0:
|
|
74
|
+
#elif prediction.size == 0 and target.size != 0:
|
|
75
|
+
elif pred.nb_object == 0 and targ.nb_object != 0:
|
|
77
76
|
tp, fp, fn, _ = 0, 0, targ.nb_object, None
|
|
78
77
|
else:
|
|
79
|
-
|
|
78
|
+
tp, fp, fn, _ = torch.nan, torch.nan, torch.nan, None
|
|
80
79
|
|
|
81
|
-
|
|
82
|
-
self.fp
|
|
83
|
-
|
|
84
|
-
|
|
85
|
-
|
|
80
|
+
# NaN because of no tn in detection
|
|
81
|
+
self.stats.append(torch.tensor([[[tp, fp, torch.nan, fn, torch.nan]]])) # fix tn and support to nan
|
|
82
|
+
|
|
83
|
+
def global_micro_compute(self) -> Tensor:
|
|
84
|
+
"""Compute metric with global/micro averagging."""
|
|
85
|
+
n_samples = len(self.stats)
|
|
86
|
+
samples_stack = torch.cat(self.stats).view(n_samples, self.nc, 5) # (N, NC, 5)
|
|
87
|
+
# sum stats accross samples
|
|
88
|
+
samples_stack = samples_stack.sum(dim=0) # (NC, 5)
|
|
89
|
+
# sum stats accross classes
|
|
90
|
+
micro_stack = torch.sum(samples_stack, dim=0) # (5,)
|
|
91
|
+
# compute metric
|
|
92
|
+
tp, fp, tn, fn, _ = micro_stack.unbind(0)
|
|
93
|
+
return self.func(tp, fp, tn, fn)
|
|
94
|
+
|
|
95
|
+
def samplewise_micro(self) -> Tensor:
|
|
96
|
+
"""Compute metric with samplewise/micro averagging."""
|
|
97
|
+
n_samples = len(self.stats)
|
|
98
|
+
samples_stack = torch.cat(self.stats).view(n_samples, self.nc, 5) # (N, NC, 5))
|
|
99
|
+
# sum stat accross classes
|
|
100
|
+
samples_stack = samples_stack.sum(dim=1) # (NC, 5)
|
|
101
|
+
# compute metric/sample
|
|
102
|
+
tp, fp, tn, fn, _ = samples_stack.unbind(dim=1)
|
|
103
|
+
samples_metrics = self.func(tp, fp, tn, fn)
|
|
104
|
+
# mean accross samples
|
|
105
|
+
return torch.nanmean(samples_metrics)
|
|
106
|
+
|
|
107
|
+
def compute_last_sample(self):
|
|
108
|
+
"""Return metrics values of the last sample in self.stats.
|
|
109
|
+
Used in combination with self.update
|
|
110
|
+
|
|
111
|
+
Returns:
|
|
112
|
+
Dict[str, Float]: dictionnary with metric value for all classes combined and for each class
|
|
113
|
+
"""
|
|
86
114
|
|
|
115
|
+
if not self.stats:
|
|
116
|
+
return {'all_cls': torch.tensor(torch.nan)}
|
|
117
|
+
|
|
118
|
+
metric_sample_dict = {}
|
|
119
|
+
|
|
120
|
+
matrix_stats_img = self.stats[-1].squeeze(0) # matrix with stats (tp, fp, tn, fn) of one sample (one row for each class)
|
|
121
|
+
nb_class = matrix_stats_img.size(dim=0)
|
|
122
|
+
tp, fp, tn, fn, _ = matrix_stats_img.nansum(dim=0).unbind(dim=0) # sum over all classes
|
|
123
|
+
all_class_metric_value = round(self.func(tp, fp, tn, fn).item(),3) # metric compute for all classes
|
|
124
|
+
metric_sample_dict['all_cls'] = all_class_metric_value
|
|
125
|
+
|
|
126
|
+
if nb_class > 1:
|
|
127
|
+
for index_class in range (nb_class) : # metric compute for each class
|
|
128
|
+
matrix_stats_img_class = matrix_stats_img[index_class,:]
|
|
129
|
+
tp, fp, tn, fn, _ = matrix_stats_img_class.unbind(dim=0) # stats (tp, fp, tn, fn) of the given class
|
|
130
|
+
class_metric_value = round(self.func(tp, fp, tn, fn).item(),3) # metric compute for the given class
|
|
131
|
+
metric_sample_dict["cls_{0}".format(index_class)] = class_metric_value
|
|
132
|
+
|
|
133
|
+
return metric_sample_dict
|
|
134
|
+
|
|
87
135
|
def compute(self) -> Dict[str, Tensor]:
|
|
88
136
|
"""Return metric computed with internal state.
|
|
89
137
|
|
|
90
138
|
Returns:
|
|
91
139
|
Dict[str, Tensor]: dictionnary with aggregation_method: value
|
|
92
140
|
"""
|
|
93
|
-
|
|
94
|
-
|
|
141
|
+
if not self.stats:
|
|
142
|
+
return {self.name: torch.tensor(torch.nan)}
|
|
143
|
+
|
|
144
|
+
# global micro
|
|
145
|
+
global_value = self.global_micro_compute()
|
|
146
|
+
# samplewise micro
|
|
147
|
+
samplewise_value = self.samplewise_micro()
|
|
148
|
+
|
|
95
149
|
metric_dict = {"global": global_value, "samplewise": samplewise_value}
|
|
96
150
|
|
|
97
151
|
return metric_dict
|
|
@@ -122,6 +176,7 @@ class ClassWiseDetectMetric(Metric):
|
|
|
122
176
|
Configuration().data_type != "semantic_mask"
|
|
123
177
|
), "Can't use detection metrics with Configuration().data_type = semantic_mask. Must be instance_mask or bbox"
|
|
124
178
|
num_classes = Configuration().num_classes
|
|
179
|
+
self.nc = num_classes
|
|
125
180
|
assert (
|
|
126
181
|
num_classes >= 2
|
|
127
182
|
), f"{num_classes} is an invalid number of classes for ClassWiseMetric (must be >= 2). If you have one class please use DetectMetric instead of ClasswiseMetric"
|
|
@@ -135,6 +190,8 @@ class ClassWiseDetectMetric(Metric):
|
|
|
135
190
|
DetectMetric(func, name=f"{name}/cls_{i}").to(self.device)
|
|
136
191
|
for i in range(num_classes)
|
|
137
192
|
]
|
|
193
|
+
self.add_state("stats", default=[], dist_reduce_fx="cat")
|
|
194
|
+
self.func = func
|
|
138
195
|
|
|
139
196
|
def update(
|
|
140
197
|
self,
|
|
@@ -155,13 +212,38 @@ class ClassWiseDetectMetric(Metric):
|
|
|
155
212
|
target = (
|
|
156
213
|
target if isinstance(target, BatchedFormat) else BatchedFormat([target])
|
|
157
214
|
)
|
|
215
|
+
list_met_class = []
|
|
158
216
|
for i, met in enumerate(self.classmetrics):
|
|
159
|
-
|
|
160
|
-
filtered_targ = self._filter_classes(target, i)
|
|
217
|
+
met_samples_result = []
|
|
161
218
|
if i == 0:
|
|
162
219
|
met.update(prediction, target)
|
|
163
220
|
else:
|
|
221
|
+
filtered_pred = self._filter_classes(prediction, i-1) # -1 because classes start at 0
|
|
222
|
+
filtered_targ = self._filter_classes(target, i-1)
|
|
164
223
|
met.update(filtered_pred, filtered_targ)
|
|
224
|
+
|
|
225
|
+
# retrieve and save in a list the stats for each sample in the batch for the given class
|
|
226
|
+
batch_size = prediction.size
|
|
227
|
+
for num_elem in range (batch_size,0,-1):
|
|
228
|
+
met_samples_result.append(met.stats[-num_elem])
|
|
229
|
+
|
|
230
|
+
list_met_class.append(met_samples_result) # save the list in the list containing the list of each class
|
|
231
|
+
|
|
232
|
+
# save the batch sample stats for each class in a matrix (N, NC, 5) and append in metric 'stats' attribute
|
|
233
|
+
for num_sample in range(batch_size):
|
|
234
|
+
sample_stats = []
|
|
235
|
+
for num_class in range(1,len(list_met_class)):
|
|
236
|
+
sample_stats.append(list_met_class[num_class][num_sample])
|
|
237
|
+
|
|
238
|
+
class_stack = torch.cat(sample_stats).view(1, Configuration().num_classes, 5) # (N, NC, 5)
|
|
239
|
+
self.stats.append(class_stack)
|
|
240
|
+
|
|
241
|
+
|
|
242
|
+
|
|
243
|
+
# for sample_list_met_class in list_met_class:
|
|
244
|
+
# sample_list_met_class = sample_list_met_class[1:]
|
|
245
|
+
# class_stack = torch.cat(sample_list_met_class).view(1, Configuration().num_classes, 5) # (N, NC, 5)
|
|
246
|
+
# self.stats.append(class_stack)
|
|
165
247
|
|
|
166
248
|
def _filter_classes(self, batchedformat: BatchedFormat, cls: int):
|
|
167
249
|
new_formats = []
|
|
@@ -170,14 +252,116 @@ class ClassWiseDetectMetric(Metric):
|
|
|
170
252
|
new_formats.append(new_form)
|
|
171
253
|
return BatchedFormat(new_formats)
|
|
172
254
|
|
|
255
|
+
def global_micro_compute(self) -> Tensor:
|
|
256
|
+
"""Compute metric with global/micro averagging."""
|
|
257
|
+
n_samples = len(self.stats)
|
|
258
|
+
samples_stack = torch.cat(self.stats).view(n_samples, self.nc, 5) # (N, NC, 5)
|
|
259
|
+
# sum stats accross samples
|
|
260
|
+
samples_stack = samples_stack.nansum(dim=0) # (NC, 5)
|
|
261
|
+
# sum stats accross classes
|
|
262
|
+
micro_stack = torch.nansum(samples_stack, dim=0) # (5,)
|
|
263
|
+
# compute metric
|
|
264
|
+
tp, fp, tn, fn, _ = micro_stack.unbind(0)
|
|
265
|
+
return self.func(tp, fp, tn, fn)
|
|
266
|
+
|
|
267
|
+
def global_macro_compute(self) -> Tuple[Tensor, Tensor]:
|
|
268
|
+
"""Compute metric with global/macro averraging.
|
|
269
|
+
Return also metric/class tensor."""
|
|
270
|
+
n_samples = len(self.stats)
|
|
271
|
+
samples_stack = torch.cat(self.stats).view(n_samples, self.nc, 5) # (N, NC, 5)
|
|
272
|
+
# sum stats accross samples
|
|
273
|
+
samples_stack = samples_stack.nansum(dim=0) # (NC, 5)
|
|
274
|
+
# compute metric/class
|
|
275
|
+
tp, fp, tn, fn, _ = samples_stack.unbind(1)
|
|
276
|
+
class_metrics = self.func(tp, fp, tn, fn) # (NC,)
|
|
277
|
+
return torch.nanmean(class_metrics), class_metrics
|
|
278
|
+
|
|
279
|
+
def samplewise_micro(self) -> Tensor:
|
|
280
|
+
"""Compute metric with samplewise/micro averagging."""
|
|
281
|
+
n_samples = len(self.stats)
|
|
282
|
+
samples_stack = torch.cat(self.stats).view(n_samples, self.nc, 5) # (N, NC, 5))
|
|
283
|
+
# sum stat accross classes
|
|
284
|
+
samples_stack = samples_stack.nansum(dim=1) # (N, 5)
|
|
285
|
+
# compute metric/sample
|
|
286
|
+
tp, fp, tn, fn, _ = samples_stack.unbind(dim=1)
|
|
287
|
+
samples_metrics = self.func(tp, fp, tn, fn)
|
|
288
|
+
# mean accross samples
|
|
289
|
+
return torch.nanmean(samples_metrics)
|
|
290
|
+
|
|
291
|
+
def samplewise_macro(self) -> Tensor:
|
|
292
|
+
"""Compute metric with samplewise/macro averagging."""
|
|
293
|
+
n_samples = len(self.stats)
|
|
294
|
+
samples_stack = torch.cat(self.stats).view(n_samples, self.nc, 5) # (N, NC, 5)
|
|
295
|
+
# compute metric/class/sample
|
|
296
|
+
tp, fp, tn, fn, _ = samples_stack.unbind(2)
|
|
297
|
+
class_metrics = self.func(tp, fp, tn, fn) # (N,NC)
|
|
298
|
+
# mean accross classes
|
|
299
|
+
macro = torch.nanmean(class_metrics, dim=1) # (N,)
|
|
300
|
+
# mean accross samples
|
|
301
|
+
macro_samplewise = torch.nanmean(macro, dim=0)
|
|
302
|
+
|
|
303
|
+
return macro_samplewise
|
|
304
|
+
|
|
305
|
+
def compute_last_sample(self):
|
|
306
|
+
"""Return metrics values of the last sample in self.stats.
|
|
307
|
+
Used in combination with self.update
|
|
308
|
+
|
|
309
|
+
Returns:
|
|
310
|
+
Dict[str, Float]: dictionnary with metric value for all classes combined and for each class
|
|
311
|
+
"""
|
|
312
|
+
|
|
313
|
+
if not self.stats:
|
|
314
|
+
return {'all_cls': torch.tensor(torch.nan)}
|
|
315
|
+
|
|
316
|
+
metric_sample_dict = {}
|
|
317
|
+
|
|
318
|
+
matrix_stats_img = self.stats[-1].squeeze(0) # matrix with stats (tp, fp, tn, fn) of one sample (one row for each class)
|
|
319
|
+
nb_class = matrix_stats_img.size(dim=0)
|
|
320
|
+
tp, fp, tn, fn, _ = matrix_stats_img.nansum(dim=0).unbind(dim=0) # sum over all classes
|
|
321
|
+
all_class_metric_value = round(self.func(tp, fp, tn, fn).item(),3) # metric compute for all classes
|
|
322
|
+
metric_sample_dict['all_cls'] = all_class_metric_value
|
|
323
|
+
|
|
324
|
+
if nb_class > 1:
|
|
325
|
+
for index_class in range (nb_class) : # metric compute for each class
|
|
326
|
+
matrix_stats_img_class = matrix_stats_img[index_class,:]
|
|
327
|
+
tp, fp, tn, fn, _ = matrix_stats_img_class.unbind(dim=0) # stats (tp, fp, tn, fn) of the given class
|
|
328
|
+
class_metric_value = round(self.func(tp, fp, tn, fn).item(),3) # metric compute for the given class
|
|
329
|
+
metric_sample_dict["cls_{0}".format(index_class)] = class_metric_value
|
|
330
|
+
|
|
331
|
+
return metric_sample_dict
|
|
332
|
+
|
|
173
333
|
def compute(self):
|
|
174
334
|
"""Return metrics values.
|
|
175
335
|
|
|
176
336
|
Returns:
|
|
177
337
|
Dict[str, Tensor]: dictionnary with all "global" DetectMetric in self.classmetrics
|
|
178
338
|
"""
|
|
179
|
-
classwisedict = {met.name: met.compute()["global"] for met in self.classmetrics}
|
|
180
|
-
return classwisedict
|
|
339
|
+
#classwisedict = {met.name: met.compute()["global"] for met in self.classmetrics}
|
|
340
|
+
#return classwisedict
|
|
341
|
+
|
|
342
|
+
if not self.stats:
|
|
343
|
+
return {self.name: torch.tensor(torch.nan)}
|
|
344
|
+
|
|
345
|
+
metric_dict = {}
|
|
346
|
+
# global micro
|
|
347
|
+
global_micro = self.global_micro_compute()
|
|
348
|
+
metric_dict.update({"_global_micro": global_micro})
|
|
349
|
+
# global macro
|
|
350
|
+
global_macro, class_metrics = self.global_macro_compute()
|
|
351
|
+
classes_dict = {
|
|
352
|
+
f"/cls_{i}": class_metrics[i]
|
|
353
|
+
for i in range(class_metrics.nelement())
|
|
354
|
+
}
|
|
355
|
+
metric_dict.update({"_global_macro": global_macro})
|
|
356
|
+
metric_dict.update(classes_dict)
|
|
357
|
+
# samplewise micro
|
|
358
|
+
samplewise_micro = self.samplewise_micro()
|
|
359
|
+
metric_dict.update({"_samplewise_micro": samplewise_micro})
|
|
360
|
+
# samplewise macro
|
|
361
|
+
samplewise_macro = self.samplewise_macro()
|
|
362
|
+
metric_dict.update({"_samplewise_macro": samplewise_macro})
|
|
363
|
+
|
|
364
|
+
return metric_dict
|
|
181
365
|
|
|
182
366
|
# override
|
|
183
367
|
def reset(self):
|
|
@@ -320,7 +504,35 @@ class ClassifMetric(Metric):
|
|
|
320
504
|
macro_samplewise = torch.nanmean(macro, dim=0)
|
|
321
505
|
|
|
322
506
|
return macro_samplewise
|
|
507
|
+
|
|
508
|
+
def compute_last_sample(self):
|
|
509
|
+
"""Return metrics values of the last sample in self.stats.
|
|
510
|
+
Used in combination with self.update
|
|
511
|
+
|
|
512
|
+
Returns:
|
|
513
|
+
Dict[str, Float]: dictionnary with metric value for all classes combined and for each class
|
|
514
|
+
"""
|
|
323
515
|
|
|
516
|
+
if not self.stats:
|
|
517
|
+
return {'all_cls': torch.tensor(torch.nan)}
|
|
518
|
+
|
|
519
|
+
metric_sample_dict = {}
|
|
520
|
+
|
|
521
|
+
matrix_stats_img = self.stats[-1].squeeze(0) # matrix with stats (tp, fp, tn, fn) of one sample (one row for each class)
|
|
522
|
+
nb_class = matrix_stats_img.size(dim=0)
|
|
523
|
+
tp, fp, tn, fn, _ = matrix_stats_img.nansum(dim=0).unbind(dim=0) # sum over all classes
|
|
524
|
+
all_class_metric_value = round(self.func(tp, fp, tn, fn).item(),3) # metric compute for all classes
|
|
525
|
+
metric_sample_dict['all_cls'] = all_class_metric_value
|
|
526
|
+
|
|
527
|
+
if nb_class > 1:
|
|
528
|
+
for index_class in range (nb_class) : # metric compute for each class
|
|
529
|
+
matrix_stats_img_class = matrix_stats_img[index_class,:]
|
|
530
|
+
tp, fp, tn, fn, _ = matrix_stats_img_class.unbind(dim=0) # stats (tp, fp, tn, fn) of the given class
|
|
531
|
+
class_metric_value = round(self.func(tp, fp, tn, fn).item(),3) # metric compute for the given class
|
|
532
|
+
metric_sample_dict["cls_{0}".format(index_class)] = class_metric_value
|
|
533
|
+
|
|
534
|
+
return metric_sample_dict
|
|
535
|
+
|
|
324
536
|
def compute(self):
|
|
325
537
|
"""Comput metric with all averag strategy and return a dict with all values."""
|
|
326
538
|
if not self.stats:
|
|
@@ -0,0 +1,36 @@
|
|
|
1
|
+
import torch
|
|
2
|
+
from torch import Tensor
|
|
3
|
+
|
|
4
|
+
|
|
5
|
+
# functionnals metrics
|
|
6
|
+
def f1score(tp: Tensor, fp: Tensor, tn: Tensor, fn: Tensor) -> Tensor:
|
|
7
|
+
for t in [tp,fp,tn,fn]:
|
|
8
|
+
assert torch.is_tensor(t), "argument must be a tensor"
|
|
9
|
+
return (2 * tp) / (2 * tp + fp + fn)
|
|
10
|
+
|
|
11
|
+
|
|
12
|
+
def precision(tp: Tensor, fp: Tensor, tn: Tensor, fn: Tensor) -> Tensor:
|
|
13
|
+
for t in [tp,fp,tn,fn]:
|
|
14
|
+
assert torch.is_tensor(t), "argument must be a tensor"
|
|
15
|
+
return (tp) / (tp + fp)
|
|
16
|
+
|
|
17
|
+
|
|
18
|
+
def recall(tp: Tensor, fp: Tensor, tn: Tensor, fn: Tensor) -> Tensor:
|
|
19
|
+
for t in [tp,fp,tn,fn]:
|
|
20
|
+
assert torch.is_tensor(t), "argument must be a tensor"
|
|
21
|
+
return (tp) / (tp + fn)
|
|
22
|
+
|
|
23
|
+
|
|
24
|
+
def iou(tp: Tensor, fp: Tensor, tn: Tensor, fn: Tensor) -> Tensor:
|
|
25
|
+
"""Compute IoU from statistics."""
|
|
26
|
+
for t in [tp,fp,tn,fn]:
|
|
27
|
+
assert torch.is_tensor(t), "argument must be a tensor"
|
|
28
|
+
return tp / (tp + fp + fn)
|
|
29
|
+
|
|
30
|
+
|
|
31
|
+
def accuracy(tp: Tensor, fp: Tensor, tn: Tensor, fn: Tensor) -> Tensor:
|
|
32
|
+
"""Compute accuracy from statistics. In case of detection, tn is nan : -> 0 for computation"""
|
|
33
|
+
for t in [tp,fp,tn,fn]:
|
|
34
|
+
assert torch.is_tensor(t), "argument must be a tensor"
|
|
35
|
+
tn[tn != tn] = 0 # set all NaN to 0 in tn
|
|
36
|
+
return (tp + tn) / (tn + tp + fp + fn)
|
|
@@ -1,8 +1,8 @@
|
|
|
1
|
-
Metadata-Version: 2.
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
2
|
Name: deepvisiontools
|
|
3
|
-
Version: 0.0
|
|
4
|
-
Summary: Overlay of PyTorch to generalize trainning and inference processes for detection
|
|
5
|
-
Author-email: Lucas Bernigaud Samatan <lucas.bernigaud-samatan@inrae.fr>, Jordan Bernigaud Samatan <jordan.bernigaud-samatan@inrae.fr>
|
|
3
|
+
Version: 0.1.0
|
|
4
|
+
Summary: Overlay of PyTorch to generalize trainning and inference processes for detection, instance segmentation and semantic segmentation tasks.
|
|
5
|
+
Author-email: Lucas Bernigaud Samatan <lucas.bernigaud-samatan@inrae.fr>, Jordan Bernigaud Samatan <jordan.bernigaud-samatan@inrae.fr>, Paul Faucher <paul.faucher@inrae.fr>
|
|
6
6
|
Maintainer-email: Jordan Bernigaud Samatan <jordan.bernigaud-samatan@inrae.fr>
|
|
7
7
|
License: GNU AFFERO GENERAL PUBLIC LICENSE
|
|
8
8
|
Version 3, 19 November 2007
|
|
@@ -670,20 +670,23 @@ Project-URL: Homepage, https://forgemia.inra.fr/ue-apc/librairies/python/deepvis
|
|
|
670
670
|
Project-URL: Issues, https://forgemia.inra.fr/ue-apc/librairies/python/deepvisiontools_v1/-/issues
|
|
671
671
|
Project-URL: Documentation, https://deepvisiontools.readthedocs.io/en/latest/
|
|
672
672
|
Project-URL: Repository, https://forgemia.inra.fr/ue-apc/librairies/python/deepvisiontools_v1
|
|
673
|
-
Classifier: Development Status ::
|
|
673
|
+
Classifier: Development Status :: 4 - Beta
|
|
674
674
|
Classifier: Programming Language :: Python :: 3.11
|
|
675
|
-
|
|
675
|
+
Classifier: Programming Language :: Python :: 3.12
|
|
676
|
+
Requires-Python: <=3.12.9,>=3.11.9
|
|
676
677
|
Description-Content-Type: text/markdown
|
|
677
678
|
License-File: LICENSE.txt
|
|
678
|
-
Requires-Dist: torch
|
|
679
|
-
Requires-Dist: torchvision
|
|
679
|
+
Requires-Dist: torch<=2.6.0,>=2.4.0
|
|
680
|
+
Requires-Dist: torchvision<=0.21.0,>=0.19.0
|
|
680
681
|
Requires-Dist: pycocotools==2.0.8
|
|
681
|
-
Requires-Dist: torchmetrics
|
|
682
|
-
Requires-Dist: tensorboard
|
|
683
|
-
Requires-Dist: ultralytics
|
|
684
|
-
Requires-Dist: transformers
|
|
682
|
+
Requires-Dist: torchmetrics<=1.7.0,>=1.4.0
|
|
683
|
+
Requires-Dist: tensorboard<=2.19.0,>=2.18.0
|
|
684
|
+
Requires-Dist: ultralytics<=8.3.95,>=8.3.28
|
|
685
|
+
Requires-Dist: transformers<=4.50.0,>=4.44.2
|
|
685
686
|
Requires-Dist: colorcet==3.1.0
|
|
686
|
-
Requires-Dist: segmentation-models-pytorch
|
|
687
|
+
Requires-Dist: segmentation-models-pytorch<=0.4.0,>=0.3.4
|
|
688
|
+
Requires-Dist: openpyxl==3.1.5
|
|
689
|
+
Dynamic: license-file
|
|
687
690
|
|
|
688
691
|
<!--
|
|
689
692
|
<p align="center">
|
|
@@ -713,6 +716,7 @@ Similar stands for other models.
|
|
|
713
716
|
All documentation about deepvisiontools, including tutorials, can be found at : https://deepvisiontools.readthedocs.io/en/latest/
|
|
714
717
|
|
|
715
718
|
## Installation
|
|
719
|
+
We have tested deepvisiontools for python versions between 3.11.9 and 3.12.9
|
|
716
720
|
|
|
717
721
|
```shell
|
|
718
722
|
pip install deepvisiontools
|
|
@@ -26,6 +26,7 @@ src/deepvisiontools/formats/errors.py
|
|
|
26
26
|
src/deepvisiontools/formats/formats.py
|
|
27
27
|
src/deepvisiontools/formats/utils.py
|
|
28
28
|
src/deepvisiontools/inference/__init__.py
|
|
29
|
+
src/deepvisiontools/inference/evaluator.py
|
|
29
30
|
src/deepvisiontools/inference/patchifier.py
|
|
30
31
|
src/deepvisiontools/inference/predictor.py
|
|
31
32
|
src/deepvisiontools/metrics/__init__.py
|
|
@@ -0,0 +1,10 @@
|
|
|
1
|
+
torch<=2.6.0,>=2.4.0
|
|
2
|
+
torchvision<=0.21.0,>=0.19.0
|
|
3
|
+
pycocotools==2.0.8
|
|
4
|
+
torchmetrics<=1.7.0,>=1.4.0
|
|
5
|
+
tensorboard<=2.19.0,>=2.18.0
|
|
6
|
+
ultralytics<=8.3.95,>=8.3.28
|
|
7
|
+
transformers<=4.50.0,>=4.44.2
|
|
8
|
+
colorcet==3.1.0
|
|
9
|
+
segmentation-models-pytorch<=0.4.0,>=0.3.4
|
|
10
|
+
openpyxl==3.1.5
|
|
@@ -1,34 +0,0 @@
|
|
|
1
|
-
import torch
|
|
2
|
-
from torch import Tensor
|
|
3
|
-
|
|
4
|
-
|
|
5
|
-
# functionnals metrics
|
|
6
|
-
def f1score(tp, fp, tn, fn):
|
|
7
|
-
return (2 * tp) / (2 * tp + fp + fn)
|
|
8
|
-
|
|
9
|
-
|
|
10
|
-
def precision(tp, fp, tn, fn):
|
|
11
|
-
# if tp + fp == 0:
|
|
12
|
-
# return torch.tensor(torch.nan)
|
|
13
|
-
return (tp) / (tp + fp)
|
|
14
|
-
|
|
15
|
-
|
|
16
|
-
def recall(tp, fp, tn, fn):
|
|
17
|
-
# if tp + fn == 0:
|
|
18
|
-
# return torch.tensor(torch.nan)
|
|
19
|
-
return (tp) / (tp + fn)
|
|
20
|
-
|
|
21
|
-
|
|
22
|
-
def iou(tp: Tensor, fp: Tensor, tn: Tensor, fn: Tensor) -> Tensor:
|
|
23
|
-
"""Compute IoU from statistics."""
|
|
24
|
-
# if tp + fp + fn == 0:
|
|
25
|
-
# return torch.tensor(torch.nan)
|
|
26
|
-
return tp / (tp + fp + fn)
|
|
27
|
-
|
|
28
|
-
|
|
29
|
-
def accuracy(tp: Tensor, fp: Tensor, tn: Tensor, fn: Tensor):
|
|
30
|
-
"""Compute accuracy from statistics. In case of detection, tn is none : -> 0 for computation"""
|
|
31
|
-
if tn == None:
|
|
32
|
-
return tp / (tp + fp + fn)
|
|
33
|
-
else:
|
|
34
|
-
return (tp + tn) / (tn + tp + fp + fn)
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/additional_augmentations.py
RENAMED
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/augmentation_class.py
RENAMED
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/batch_augmentations.py
RENAMED
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/data_reader/__init__.py
RENAMED
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/data/data_reader/readers.py
RENAMED
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/mask2former/__init__.py
RENAMED
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8/__init__.py
RENAMED
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8seg/__init__.py
RENAMED
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8seg/errors.py
RENAMED
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8seg/utils.py
RENAMED
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/models/yolov8seg/yolov8seg.py
RENAMED
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/preprocessing/__init__.py
RENAMED
|
File without changes
|
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools/preprocessing/preprocessing.py
RENAMED
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
{deepvisiontools-0.0.4 → deepvisiontools-0.1.0}/src/deepvisiontools.egg-info/dependency_links.txt
RENAMED
|
File without changes
|
|
File without changes
|