Merge pull request #4241 from kinaryml:interactive-segmenter-python

PiperOrigin-RevId: 522470912
This commit is contained in:
Copybara-Service 2023-04-06 16:47:46 -07:00
commit 6495f8624b
4 changed files with 609 additions and 0 deletions

View File

@ -99,6 +99,7 @@ cc_library(
"//mediapipe/tasks/cc/vision/image_classifier:image_classifier_graph", "//mediapipe/tasks/cc/vision/image_classifier:image_classifier_graph",
"//mediapipe/tasks/cc/vision/image_embedder:image_embedder_graph", "//mediapipe/tasks/cc/vision/image_embedder:image_embedder_graph",
"//mediapipe/tasks/cc/vision/image_segmenter:image_segmenter_graph", "//mediapipe/tasks/cc/vision/image_segmenter:image_segmenter_graph",
"//mediapipe/tasks/cc/vision/interactive_segmenter:interactive_segmenter_graph",
"//mediapipe/tasks/cc/vision/object_detector:object_detector_graph", "//mediapipe/tasks/cc/vision/object_detector:object_detector_graph",
], ],
) )

View File

@ -0,0 +1,332 @@
# Copyright 2023 The MediaPipe Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Tests for interactive segmenter."""
import enum
import os
from absl.testing import absltest
from absl.testing import parameterized
import cv2
import numpy as np
from mediapipe.python._framework_bindings import image as image_module
from mediapipe.python._framework_bindings import image_frame
from mediapipe.tasks.python.components.containers import keypoint as keypoint_module
from mediapipe.tasks.python.components.containers import rect
from mediapipe.tasks.python.core import base_options as base_options_module
from mediapipe.tasks.python.test import test_utils
from mediapipe.tasks.python.vision import interactive_segmenter
from mediapipe.tasks.python.vision.core import image_processing_options as image_processing_options_module
_BaseOptions = base_options_module.BaseOptions
_Image = image_module.Image
_ImageFormat = image_frame.ImageFormat
_NormalizedKeypoint = keypoint_module.NormalizedKeypoint
_Rect = rect.Rect
_OutputType = interactive_segmenter.InteractiveSegmenterOptions.OutputType
_InteractiveSegmenter = interactive_segmenter.InteractiveSegmenter
_InteractiveSegmenterOptions = interactive_segmenter.InteractiveSegmenterOptions
_RegionOfInterest = interactive_segmenter.RegionOfInterest
_Format = interactive_segmenter.RegionOfInterest.Format
_ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions
_MODEL_FILE = 'ptm_512_hdt_ptm_woid.tflite'
_CATS_AND_DOGS = 'cats_and_dogs.jpg'
_CATS_AND_DOGS_MASK_DOG_1 = 'cats_and_dogs_mask_dog1.png'
_CATS_AND_DOGS_MASK_DOG_2 = 'cats_and_dogs_mask_dog2.png'
_MASK_MAGNIFICATION_FACTOR = 255
_MASK_SIMILARITY_THRESHOLD = 0.97
_TEST_DATA_DIR = 'mediapipe/tasks/testdata/vision'
def _calculate_soft_iou(m1, m2):
intersection_sum = np.sum(m1 * m2)
union_sum = np.sum(m1 * m1) + np.sum(m2 * m2) - intersection_sum
if union_sum > 0:
return intersection_sum / union_sum
else:
return 0
def _similar_to_float_mask(actual_mask, expected_mask, similarity_threshold):
actual_mask = actual_mask.numpy_view()
expected_mask = expected_mask.numpy_view() / 255.0
return (
actual_mask.shape == expected_mask.shape
and _calculate_soft_iou(actual_mask, expected_mask) > similarity_threshold
)
def _similar_to_uint8_mask(actual_mask, expected_mask, similarity_threshold):
actual_mask_pixels = actual_mask.numpy_view().flatten()
expected_mask_pixels = expected_mask.numpy_view().flatten()
consistent_pixels = 0
num_pixels = len(expected_mask_pixels)
for index in range(num_pixels):
consistent_pixels += (
actual_mask_pixels[index] * _MASK_MAGNIFICATION_FACTOR
== expected_mask_pixels[index]
)
return consistent_pixels / num_pixels >= similarity_threshold
class ModelFileType(enum.Enum):
FILE_CONTENT = 1
FILE_NAME = 2
class InteractiveSegmenterTest(parameterized.TestCase):
def setUp(self):
super().setUp()
# Load the test input image.
self.test_image = _Image.create_from_file(
test_utils.get_test_data_path(
os.path.join(_TEST_DATA_DIR, _CATS_AND_DOGS)
)
)
# Loads ground truth segmentation file.
self.test_seg_image = self._load_segmentation_mask(
_CATS_AND_DOGS_MASK_DOG_1
)
self.model_path = test_utils.get_test_data_path(
os.path.join(_TEST_DATA_DIR, _MODEL_FILE)
)
def _load_segmentation_mask(self, file_path: str):
# Loads ground truth segmentation file.
gt_segmentation_data = cv2.imread(
test_utils.get_test_data_path(os.path.join(_TEST_DATA_DIR, file_path)),
cv2.IMREAD_GRAYSCALE,
)
return _Image(_ImageFormat.GRAY8, gt_segmentation_data)
def test_create_from_file_succeeds_with_valid_model_path(self):
# Creates with default option and valid model file successfully.
with _InteractiveSegmenter.create_from_model_path(
self.model_path
) as segmenter:
self.assertIsInstance(segmenter, _InteractiveSegmenter)
def test_create_from_options_succeeds_with_valid_model_path(self):
# Creates with options containing model file successfully.
base_options = _BaseOptions(model_asset_path=self.model_path)
options = _InteractiveSegmenterOptions(base_options=base_options)
with _InteractiveSegmenter.create_from_options(options) as segmenter:
self.assertIsInstance(segmenter, _InteractiveSegmenter)
def test_create_from_options_fails_with_invalid_model_path(self):
with self.assertRaisesRegex(
RuntimeError, 'Unable to open file at /path/to/invalid/model.tflite'
):
base_options = _BaseOptions(
model_asset_path='/path/to/invalid/model.tflite'
)
options = _InteractiveSegmenterOptions(base_options=base_options)
_InteractiveSegmenter.create_from_options(options)
def test_create_from_options_succeeds_with_valid_model_content(self):
# Creates with options containing model content successfully.
with open(self.model_path, 'rb') as f:
base_options = _BaseOptions(model_asset_buffer=f.read())
options = _InteractiveSegmenterOptions(base_options=base_options)
segmenter = _InteractiveSegmenter.create_from_options(options)
self.assertIsInstance(segmenter, _InteractiveSegmenter)
@parameterized.parameters(
(
ModelFileType.FILE_NAME,
_RegionOfInterest.Format.KEYPOINT,
_NormalizedKeypoint(0.44, 0.7),
_CATS_AND_DOGS_MASK_DOG_1,
0.84,
),
(
ModelFileType.FILE_CONTENT,
_RegionOfInterest.Format.KEYPOINT,
_NormalizedKeypoint(0.44, 0.7),
_CATS_AND_DOGS_MASK_DOG_1,
0.84,
),
(
ModelFileType.FILE_NAME,
_RegionOfInterest.Format.KEYPOINT,
_NormalizedKeypoint(0.66, 0.66),
_CATS_AND_DOGS_MASK_DOG_2,
_MASK_SIMILARITY_THRESHOLD,
),
(
ModelFileType.FILE_CONTENT,
_RegionOfInterest.Format.KEYPOINT,
_NormalizedKeypoint(0.66, 0.66),
_CATS_AND_DOGS_MASK_DOG_2,
_MASK_SIMILARITY_THRESHOLD,
),
)
def test_segment_succeeds_with_category_mask(
self,
model_file_type,
roi_format,
keypoint,
output_mask,
similarity_threshold,
):
# Creates segmenter.
if model_file_type is ModelFileType.FILE_NAME:
base_options = _BaseOptions(model_asset_path=self.model_path)
elif model_file_type is ModelFileType.FILE_CONTENT:
with open(self.model_path, 'rb') as f:
model_content = f.read()
base_options = _BaseOptions(model_asset_buffer=model_content)
else:
# Should never happen
raise ValueError('model_file_type is invalid.')
options = _InteractiveSegmenterOptions(
base_options=base_options, output_type=_OutputType.CATEGORY_MASK
)
segmenter = _InteractiveSegmenter.create_from_options(options)
# Performs image segmentation on the input.
roi = _RegionOfInterest(format=roi_format, keypoint=keypoint)
category_masks = segmenter.segment(self.test_image, roi)
self.assertLen(category_masks, 1)
category_mask = category_masks[0]
result_pixels = category_mask.numpy_view().flatten()
# Check if data type of `category_mask` is correct.
self.assertEqual(result_pixels.dtype, np.uint8)
# Loads ground truth segmentation file.
test_seg_image = self._load_segmentation_mask(output_mask)
self.assertTrue(
_similar_to_uint8_mask(
category_masks[0], test_seg_image, similarity_threshold
),
(
'Number of pixels in the candidate mask differing from that of the'
f' ground truth mask exceeds {similarity_threshold}.'
),
)
# Closes the segmenter explicitly when the segmenter is not used in
# a context.
segmenter.close()
@parameterized.parameters(
(
_RegionOfInterest.Format.KEYPOINT,
_NormalizedKeypoint(0.44, 0.7),
_CATS_AND_DOGS_MASK_DOG_1,
0.84,
),
(
_RegionOfInterest.Format.KEYPOINT,
_NormalizedKeypoint(0.66, 0.66),
_CATS_AND_DOGS_MASK_DOG_2,
_MASK_SIMILARITY_THRESHOLD,
),
)
def test_segment_succeeds_with_confidence_mask(
self, roi_format, keypoint, output_mask, similarity_threshold
):
# Creates segmenter.
base_options = _BaseOptions(model_asset_path=self.model_path)
roi = _RegionOfInterest(format=roi_format, keypoint=keypoint)
# Run segmentation on the model in CONFIDENCE_MASK mode.
options = _InteractiveSegmenterOptions(
base_options=base_options, output_type=_OutputType.CONFIDENCE_MASK
)
with _InteractiveSegmenter.create_from_options(options) as segmenter:
# Perform segmentation
confidence_masks = segmenter.segment(self.test_image, roi)
# Check if confidence mask shape is correct.
self.assertLen(
confidence_masks,
2,
'Number of confidence masks must match with number of categories.',
)
# Loads ground truth segmentation file.
expected_mask = self._load_segmentation_mask(output_mask)
self.assertTrue(
_similar_to_float_mask(
confidence_masks[1], expected_mask, similarity_threshold
)
)
def test_segment_succeeds_with_rotation(self):
# Creates segmenter.
base_options = _BaseOptions(model_asset_path=self.model_path)
roi = _RegionOfInterest(
format=_RegionOfInterest.Format.KEYPOINT,
keypoint=_NormalizedKeypoint(0.66, 0.66),
)
# Run segmentation on the model in CONFIDENCE_MASK mode.
options = _InteractiveSegmenterOptions(
base_options=base_options, output_type=_OutputType.CONFIDENCE_MASK
)
with _InteractiveSegmenter.create_from_options(options) as segmenter:
# Perform segmentation
image_processing_options = _ImageProcessingOptions(rotation_degrees=-90)
confidence_masks = segmenter.segment(
self.test_image, roi, image_processing_options
)
# Check if confidence mask shape is correct.
self.assertLen(
confidence_masks,
2,
'Number of confidence masks must match with number of categories.',
)
def test_segment_fails_with_roi_in_image_processing_options(self):
# Creates segmenter.
base_options = _BaseOptions(model_asset_path=self.model_path)
roi = _RegionOfInterest(
format=_RegionOfInterest.Format.KEYPOINT,
keypoint=_NormalizedKeypoint(0.66, 0.66),
)
# Run segmentation on the model in CONFIDENCE_MASK mode.
options = _InteractiveSegmenterOptions(
base_options=base_options, output_type=_OutputType.CONFIDENCE_MASK
)
with self.assertRaisesRegex(
ValueError, "This task doesn't support region-of-interest."
):
with _InteractiveSegmenter.create_from_options(options) as segmenter:
# Perform segmentation
image_processing_options = _ImageProcessingOptions(
_Rect(left=0.1, top=0, right=0.9, bottom=1)
)
segmenter.segment(self.test_image, roi, image_processing_options)
if __name__ == '__main__':
absltest.main()

View File

@ -106,6 +106,28 @@ py_library(
], ],
) )
py_library(
name = "interactive_segmenter",
srcs = [
"interactive_segmenter.py",
],
deps = [
"//mediapipe/python:_framework_bindings",
"//mediapipe/python:packet_creator",
"//mediapipe/python:packet_getter",
"//mediapipe/tasks/cc/vision/image_segmenter/proto:image_segmenter_graph_options_py_pb2",
"//mediapipe/tasks/cc/vision/image_segmenter/proto:segmenter_options_py_pb2",
"//mediapipe/tasks/python/components/containers:keypoint",
"//mediapipe/tasks/python/core:base_options",
"//mediapipe/tasks/python/core:optional_dependencies",
"//mediapipe/tasks/python/core:task_info",
"//mediapipe/tasks/python/vision/core:base_vision_task_api",
"//mediapipe/tasks/python/vision/core:image_processing_options",
"//mediapipe/tasks/python/vision/core:vision_task_running_mode",
"//mediapipe/util:render_data_py_pb2",
],
)
py_library( py_library(
name = "gesture_recognizer", name = "gesture_recognizer",
srcs = [ srcs = [

View File

@ -0,0 +1,254 @@
# Copyright 2023 The MediaPipe Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""MediaPipe interactive segmenter task."""
import dataclasses
import enum
from typing import List, Optional
from mediapipe.python import packet_creator
from mediapipe.python import packet_getter
from mediapipe.python._framework_bindings import image as image_module
from mediapipe.tasks.cc.vision.image_segmenter.proto import image_segmenter_graph_options_pb2
from mediapipe.tasks.cc.vision.image_segmenter.proto import segmenter_options_pb2
from mediapipe.tasks.python.components.containers import keypoint as keypoint_module
from mediapipe.tasks.python.core import base_options as base_options_module
from mediapipe.tasks.python.core import task_info as task_info_module
from mediapipe.tasks.python.core.optional_dependencies import doc_controls
from mediapipe.tasks.python.vision.core import base_vision_task_api
from mediapipe.tasks.python.vision.core import image_processing_options as image_processing_options_module
from mediapipe.tasks.python.vision.core import vision_task_running_mode
from mediapipe.util import render_data_pb2
_BaseOptions = base_options_module.BaseOptions
_RenderDataProto = render_data_pb2.RenderData
_SegmenterOptionsProto = segmenter_options_pb2.SegmenterOptions
_ImageSegmenterGraphOptionsProto = (
image_segmenter_graph_options_pb2.ImageSegmenterGraphOptions
)
_RunningMode = vision_task_running_mode.VisionTaskRunningMode
_ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions
_TaskInfo = task_info_module.TaskInfo
_SEGMENTATION_OUT_STREAM_NAME = 'segmented_mask_out'
_SEGMENTATION_TAG = 'GROUPED_SEGMENTATION'
_IMAGE_IN_STREAM_NAME = 'image_in'
_IMAGE_OUT_STREAM_NAME = 'image_out'
_ROI_STREAM_NAME = 'roi_in'
_ROI_TAG = 'ROI'
_NORM_RECT_STREAM_NAME = 'norm_rect_in'
_NORM_RECT_TAG = 'NORM_RECT'
_IMAGE_TAG = 'IMAGE'
_TASK_GRAPH_NAME = (
'mediapipe.tasks.vision.interactive_segmenter.InteractiveSegmenterGraph'
)
@dataclasses.dataclass
class InteractiveSegmenterOptions:
"""Options for the interactive segmenter task.
Attributes:
base_options: Base options for the interactive segmenter task.
output_type: The output mask type allows specifying the type of
post-processing to perform on the raw model results.
"""
class OutputType(enum.Enum):
UNSPECIFIED = 0
CATEGORY_MASK = 1
CONFIDENCE_MASK = 2
base_options: _BaseOptions
output_type: Optional[OutputType] = OutputType.CATEGORY_MASK
@doc_controls.do_not_generate_docs
def to_pb2(self) -> _ImageSegmenterGraphOptionsProto:
"""Generates an InteractiveSegmenterOptions protobuf object."""
base_options_proto = self.base_options.to_pb2()
base_options_proto.use_stream_mode = False
segmenter_options_proto = _SegmenterOptionsProto(
output_type=self.output_type.value
)
return _ImageSegmenterGraphOptionsProto(
base_options=base_options_proto,
segmenter_options=segmenter_options_proto,
)
@dataclasses.dataclass
class RegionOfInterest:
"""The Region-Of-Interest (ROI) to interact with."""
class Format(enum.Enum):
UNSPECIFIED = 0
KEYPOINT = 1
format: Format
keypoint: Optional[keypoint_module.NormalizedKeypoint] = None
def _convert_roi_to_render_data(roi: RegionOfInterest) -> _RenderDataProto:
"""Converts region of interest to render data proto."""
result = _RenderDataProto()
if roi is not None:
if roi.format == RegionOfInterest.Format.UNSPECIFIED:
raise ValueError('RegionOfInterest format not specified.')
elif roi.format == RegionOfInterest.Format.KEYPOINT:
if roi.keypoint is not None:
annotation = result.render_annotations.add()
annotation.color.r = 255
point = annotation.point
point.normalized = True
point.x = roi.keypoint.x
point.y = roi.keypoint.y
return result
else:
raise ValueError('Please specify the Region-of-interest for segmentation.')
raise ValueError('Unrecognized format.')
class InteractiveSegmenter(base_vision_task_api.BaseVisionTaskApi):
"""Class that performs interactive segmentation on images.
Users can represent user interaction through `RegionOfInterest`, which gives
a hint to InteractiveSegmenter to perform segmentation focusing on the given
region of interest.
The API expects a TFLite model with mandatory TFLite Model Metadata.
Input tensor:
(kTfLiteUInt8/kTfLiteFloat32)
- image input of size `[batch x height x width x channels]`.
- batch inference is not supported (`batch` is required to be 1).
- RGB and greyscale inputs are supported (`channels` is required to be
1 or 3).
- if type is kTfLiteFloat32, NormalizationOptions are required to be
attached to the metadata for input normalization.
Output tensors:
(kTfLiteUInt8/kTfLiteFloat32)
- list of segmented masks.
- if `output_type` is CATEGORY_MASK, uint8 Image, Image vector of size 1.
- if `output_type` is CONFIDENCE_MASK, float32 Image list of size
`channels`.
- batch is always 1
An example of such model can be found at:
https://tfhub.dev/tensorflow/lite-model/deeplabv3/1/metadata/2
"""
@classmethod
def create_from_model_path(cls, model_path: str) -> 'InteractiveSegmenter':
"""Creates an `InteractiveSegmenter` object from a TensorFlow Lite model and the default `InteractiveSegmenterOptions`.
Note that the created `InteractiveSegmenter` instance is in image mode, for
performing image segmentation on single image inputs.
Args:
model_path: Path to the model.
Returns:
`InteractiveSegmenter` object that's created from the model file and the
default `InteractiveSegmenterOptions`.
Raises:
ValueError: If failed to create `InteractiveSegmenter` object from the
provided file such as invalid file path.
RuntimeError: If other types of error occurred.
"""
base_options = _BaseOptions(model_asset_path=model_path)
options = InteractiveSegmenterOptions(base_options=base_options)
return cls.create_from_options(options)
@classmethod
def create_from_options(
cls, options: InteractiveSegmenterOptions
) -> 'InteractiveSegmenter':
"""Creates the `InteractiveSegmenter` object from interactive segmenter options.
Args:
options: Options for the interactive segmenter task.
Returns:
`InteractiveSegmenter` object that's created from `options`.
Raises:
ValueError: If failed to create `InteractiveSegmenter` object from
`InteractiveSegmenterOptions` such as missing the model.
RuntimeError: If other types of error occurred.
"""
task_info = _TaskInfo(
task_graph=_TASK_GRAPH_NAME,
input_streams=[
':'.join([_IMAGE_TAG, _IMAGE_IN_STREAM_NAME]),
':'.join([_ROI_TAG, _ROI_STREAM_NAME]),
':'.join([_NORM_RECT_TAG, _NORM_RECT_STREAM_NAME]),
],
output_streams=[
':'.join([_SEGMENTATION_TAG, _SEGMENTATION_OUT_STREAM_NAME]),
':'.join([_IMAGE_TAG, _IMAGE_OUT_STREAM_NAME]),
],
task_options=options,
)
return cls(
task_info.generate_graph_config(enable_flow_limiting=False),
_RunningMode.IMAGE,
None,
)
def segment(
self,
image: image_module.Image,
roi: RegionOfInterest,
image_processing_options: Optional[_ImageProcessingOptions] = None,
) -> List[image_module.Image]:
"""Performs the actual segmentation task on the provided MediaPipe Image.
The image can be of any size with format RGB.
Args:
image: MediaPipe Image.
roi: Optional user-specified region of interest for segmentation.
image_processing_options: Options for image processing.
Returns:
If the output_type is CATEGORY_MASK, the returned vector of images is
per-category segmented image mask.
If the output_type is CONFIDENCE_MASK, the returned vector of images
contains only one confidence image mask. A segmentation result object that
contains a list of segmentation masks as images.
Raises:
ValueError: If any of the input arguments is invalid.
RuntimeError: If image segmentation failed to run.
"""
normalized_rect = self.convert_to_normalized_rect(
image_processing_options, image, roi_allowed=False
)
render_data_proto = _convert_roi_to_render_data(roi)
output_packets = self._process_image_data({
_IMAGE_IN_STREAM_NAME: packet_creator.create_image(image),
_ROI_STREAM_NAME: packet_creator.create_proto(render_data_proto),
_NORM_RECT_STREAM_NAME: packet_creator.create_proto(
normalized_rect.to_pb2()
),
})
segmentation_result = packet_getter.get_image_list(
output_packets[_SEGMENTATION_OUT_STREAM_NAME]
)
return segmentation_result