Added the PoseLandmarker Python API and a simple test
This commit is contained in:
kinaryml
parent
dbeb5a8126
commit
5f5ce22020
|
|
@ -92,6 +92,7 @@ cc_library(
|
|||
"//mediapipe/tasks/cc/audio/audio_embedder:audio_embedder_graph",
|
||||
"//mediapipe/tasks/cc/vision/face_detector:face_detector_graph",
|
||||
"//mediapipe/tasks/cc/vision/face_landmarker:face_landmarker_graph",
|
||||
"//mediapipe/tasks/cc/vision/pose_landmarker:pose_landmarker_graph",
|
||||
"//mediapipe/tasks/cc/vision/face_stylizer:face_stylizer_graph",
|
||||
"//mediapipe/tasks/cc/vision/gesture_recognizer:gesture_recognizer_graph",
|
||||
"//mediapipe/tasks/cc/vision/image_classifier:image_classifier_graph",
|
||||
|
|
|
|||
|
|
@ -162,3 +162,26 @@ py_test(
|
|||
"@com_google_protobuf//:protobuf_python",
|
||||
],
|
||||
)
|
||||
|
||||
py_test(
|
||||
name = "pose_landmarker_test",
|
||||
srcs = ["pose_landmarker_test.py"],
|
||||
data = [
|
||||
"//mediapipe/tasks/testdata/vision:test_images",
|
||||
"//mediapipe/tasks/testdata/vision:test_models",
|
||||
"//mediapipe/tasks/testdata/vision:test_protos",
|
||||
],
|
||||
deps = [
|
||||
"//mediapipe/tasks/cc/components/containers/proto:landmarks_detection_result_py_pb2",
|
||||
"//mediapipe/python:_framework_bindings",
|
||||
"//mediapipe/tasks/python/components/containers:landmark",
|
||||
"//mediapipe/tasks/python/components/containers:landmark_detection_result",
|
||||
"//mediapipe/tasks/python/components/containers:rect",
|
||||
"//mediapipe/tasks/python/core:base_options",
|
||||
"//mediapipe/tasks/python/test:test_utils",
|
||||
"//mediapipe/tasks/python/vision:pose_landmarker",
|
||||
"//mediapipe/tasks/python/vision/core:image_processing_options",
|
||||
"//mediapipe/tasks/python/vision/core:vision_task_running_mode",
|
||||
"@com_google_protobuf//:protobuf_python",
|
||||
],
|
||||
)
|
||||
|
|
|
|||
177
mediapipe/tasks/python/test/vision/pose_landmarker_test.py
Normal file
177
mediapipe/tasks/python/test/vision/pose_landmarker_test.py
Normal file
|
|
@ -0,0 +1,177 @@
|
|||
# Copyright 2022 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 pose landmarker."""
|
||||
|
||||
import enum
|
||||
from unittest import mock
|
||||
|
||||
from absl.testing import absltest
|
||||
from absl.testing import parameterized
|
||||
import numpy as np
|
||||
|
||||
from google.protobuf import text_format
|
||||
from mediapipe.python._framework_bindings import image as image_module
|
||||
from mediapipe.tasks.cc.components.containers.proto import landmarks_detection_result_pb2
|
||||
from mediapipe.tasks.python.components.containers import landmark as landmark_module
|
||||
from mediapipe.tasks.python.components.containers import landmark_detection_result as landmark_detection_result_module
|
||||
from mediapipe.tasks.python.components.containers import rect as rect_module
|
||||
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 pose_landmarker
|
||||
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 as running_mode_module
|
||||
|
||||
PoseLandmarkerResult = pose_landmarker.PoseLandmarkerResult
|
||||
_LandmarksDetectionResultProto = landmarks_detection_result_pb2.LandmarksDetectionResult
|
||||
_BaseOptions = base_options_module.BaseOptions
|
||||
_Rect = rect_module.Rect
|
||||
_Landmark = landmark_module.Landmark
|
||||
_NormalizedLandmark = landmark_module.NormalizedLandmark
|
||||
_LandmarksDetectionResult = landmark_detection_result_module.LandmarksDetectionResult
|
||||
_Image = image_module.Image
|
||||
_PoseLandmarker = pose_landmarker.PoseLandmarker
|
||||
_PoseLandmarkerOptions = pose_landmarker.PoseLandmarkerOptions
|
||||
_RUNNING_MODE = running_mode_module.VisionTaskRunningMode
|
||||
_ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions
|
||||
|
||||
_POSE_LANDMARKER_BUNDLE_ASSET_FILE = 'pose_landmarker.task'
|
||||
_BURGER_IMAGE = 'burger.jpg'
|
||||
_POSE_IMAGE = 'pose.jpg'
|
||||
_POSE_LANDMARKS = 'pose_landmarks.pbtxt'
|
||||
_LANDMARKS_ERROR_TOLERANCE = 0.03
|
||||
_LANDMARKS_ON_VIDEO_ERROR_TOLERANCE = 0.03
|
||||
|
||||
|
||||
def _get_expected_pose_landmarker_result(
|
||||
file_path: str) -> PoseLandmarkerResult:
|
||||
landmarks_detection_result_file_path = test_utils.get_test_data_path(
|
||||
file_path)
|
||||
with open(landmarks_detection_result_file_path, 'rb') as f:
|
||||
landmarks_detection_result_proto = _LandmarksDetectionResultProto()
|
||||
# Use this if a .pb file is available.
|
||||
# landmarks_detection_result_proto.ParseFromString(f.read())
|
||||
text_format.Parse(f.read(), landmarks_detection_result_proto)
|
||||
landmarks_detection_result = _LandmarksDetectionResult.create_from_pb2(
|
||||
landmarks_detection_result_proto)
|
||||
return PoseLandmarkerResult(
|
||||
pose_landmarks=[landmarks_detection_result.landmarks],
|
||||
pose_world_landmarks=[],
|
||||
pose_auxiliary_landmarks=[]
|
||||
)
|
||||
|
||||
|
||||
class ModelFileType(enum.Enum):
|
||||
FILE_CONTENT = 1
|
||||
FILE_NAME = 2
|
||||
|
||||
|
||||
class PoseLandmarkerTest(parameterized.TestCase):
|
||||
|
||||
def setUp(self):
|
||||
super().setUp()
|
||||
self.test_image = _Image.create_from_file(
|
||||
test_utils.get_test_data_path(_POSE_IMAGE))
|
||||
self.model_path = test_utils.get_test_data_path(
|
||||
_POSE_LANDMARKER_BUNDLE_ASSET_FILE)
|
||||
|
||||
def _expect_pose_landmarker_results_correct(
|
||||
self,
|
||||
actual_result: PoseLandmarkerResult,
|
||||
expected_result: PoseLandmarkerResult,
|
||||
error_tolerance: float
|
||||
):
|
||||
# Expects to have the same number of poses detected.
|
||||
self.assertLen(actual_result.pose_landmarks,
|
||||
len(expected_result.pose_landmarks))
|
||||
self.assertLen(actual_result.pose_world_landmarks,
|
||||
len(expected_result.pose_world_landmarks))
|
||||
self.assertLen(actual_result.pose_auxiliary_landmarks,
|
||||
len(expected_result.pose_auxiliary_landmarks))
|
||||
# Actual landmarks match expected landmarks.
|
||||
actual_landmarks = actual_result.pose_landmarks[0]
|
||||
expected_landmarks = expected_result.pose_landmarks[0]
|
||||
for i, pose_landmark in enumerate(actual_landmarks):
|
||||
self.assertAlmostEqual(
|
||||
pose_landmark.x,
|
||||
expected_landmarks[i].x,
|
||||
delta=error_tolerance
|
||||
)
|
||||
self.assertAlmostEqual(
|
||||
pose_landmark.y,
|
||||
expected_landmarks[i].y,
|
||||
delta=error_tolerance
|
||||
)
|
||||
|
||||
def test_create_from_file_succeeds_with_valid_model_path(self):
|
||||
# Creates with default option and valid model file successfully.
|
||||
with _PoseLandmarker.create_from_model_path(self.model_path) as landmarker:
|
||||
self.assertIsInstance(landmarker, _PoseLandmarker)
|
||||
|
||||
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 = _PoseLandmarkerOptions(base_options=base_options)
|
||||
with _PoseLandmarker.create_from_options(options) as landmarker:
|
||||
self.assertIsInstance(landmarker, _PoseLandmarker)
|
||||
|
||||
def test_create_from_options_fails_with_invalid_model_path(self):
|
||||
# Invalid empty model path.
|
||||
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 = _PoseLandmarkerOptions(base_options=base_options)
|
||||
_PoseLandmarker.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 = _PoseLandmarkerOptions(base_options=base_options)
|
||||
landmarker = _PoseLandmarker.create_from_options(options)
|
||||
self.assertIsInstance(landmarker, _PoseLandmarker)
|
||||
|
||||
@parameterized.parameters(
|
||||
(ModelFileType.FILE_NAME,
|
||||
_get_expected_pose_landmarker_result(_POSE_LANDMARKS)),
|
||||
(ModelFileType.FILE_CONTENT,
|
||||
_get_expected_pose_landmarker_result(_POSE_LANDMARKS)))
|
||||
def test_detect(self, model_file_type, expected_detection_result):
|
||||
# Creates pose landmarker.
|
||||
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 = _PoseLandmarkerOptions(base_options=base_options)
|
||||
landmarker = _PoseLandmarker.create_from_options(options)
|
||||
|
||||
# Performs pose landmarks detection on the input.
|
||||
detection_result = landmarker.detect(self.test_image)
|
||||
# Comparing results.
|
||||
self._expect_pose_landmarker_results_correct(
|
||||
detection_result, expected_detection_result, _LANDMARKS_ERROR_TOLERANCE
|
||||
)
|
||||
# Closes the pose landmarker explicitly when the pose landmarker is not used
|
||||
# in a context.
|
||||
landmarker.close()
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
absltest.main()
|
||||
|
|
@ -179,6 +179,27 @@ py_library(
|
|||
],
|
||||
)
|
||||
|
||||
py_library(
|
||||
name = "pose_landmarker",
|
||||
srcs = [
|
||||
"pose_landmarker.py",
|
||||
],
|
||||
deps = [
|
||||
"//mediapipe/framework/formats:landmark_py_pb2",
|
||||
"//mediapipe/python:_framework_bindings",
|
||||
"//mediapipe/python:packet_creator",
|
||||
"//mediapipe/python:packet_getter",
|
||||
"//mediapipe/tasks/cc/vision/pose_landmarker/proto:pose_landmarker_graph_options_py_pb2",
|
||||
"//mediapipe/tasks/python/components/containers:landmark",
|
||||
"//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",
|
||||
],
|
||||
)
|
||||
|
||||
py_library(
|
||||
name = "face_detector",
|
||||
srcs = [
|
||||
|
|
|
|||
428
mediapipe/tasks/python/vision/pose_landmarker.py
Normal file
428
mediapipe/tasks/python/vision/pose_landmarker.py
Normal file
|
|
@ -0,0 +1,428 @@
|
|||
# Copyright 2022 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 pose landmarker task."""
|
||||
|
||||
import dataclasses
|
||||
from typing import Callable, Mapping, Optional, List
|
||||
|
||||
from mediapipe.framework.formats import landmark_pb2
|
||||
from mediapipe.python import packet_creator
|
||||
from mediapipe.python import packet_getter
|
||||
from mediapipe.python._framework_bindings import image as image_module
|
||||
from mediapipe.python._framework_bindings import packet as packet_module
|
||||
from mediapipe.tasks.cc.vision.pose_landmarker.proto import pose_landmarker_graph_options_pb2
|
||||
from mediapipe.tasks.python.components.containers import landmark as landmark_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 as running_mode_module
|
||||
|
||||
_BaseOptions = base_options_module.BaseOptions
|
||||
_PoseLandmarkerGraphOptionsProto = (
|
||||
pose_landmarker_graph_options_pb2.PoseLandmarkerGraphOptions
|
||||
)
|
||||
_RunningMode = running_mode_module.VisionTaskRunningMode
|
||||
_ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions
|
||||
_TaskInfo = task_info_module.TaskInfo
|
||||
|
||||
_IMAGE_IN_STREAM_NAME = 'image_in'
|
||||
_IMAGE_OUT_STREAM_NAME = 'image_out'
|
||||
_IMAGE_TAG = 'IMAGE'
|
||||
_NORM_RECT_STREAM_NAME = 'norm_rect_in'
|
||||
_NORM_RECT_TAG = 'NORM_RECT'
|
||||
_SEGMENTATION_MASK_STREAM_NAME = 'segmentation_mask'
|
||||
_SEGMENTATION_MASK_TAG = 'SEGMENTATION_MASK'
|
||||
_NORM_LANDMARKS_STREAM_NAME = 'norm_landmarks'
|
||||
_NORM_LANDMARKS_TAG = 'NORM_LANDMARKS'
|
||||
_POSE_WORLD_LANDMARKS_STREAM_NAME = 'world_landmarks'
|
||||
_POSE_WORLD_LANDMARKS_TAG = 'WORLD_LANDMARKS'
|
||||
_POSE_AUXILIARY_LANDMARKS_STREAM_NAME = 'auxiliary_landmarks'
|
||||
_POSE_AUXILIARY_LANDMARKS_TAG = 'AUXILIARY_LANDMARKS'
|
||||
_TASK_GRAPH_NAME = 'mediapipe.tasks.vision.pose_landmarker.PoseLandmarkerGraph'
|
||||
_MICRO_SECONDS_PER_MILLISECOND = 1000
|
||||
|
||||
|
||||
@dataclasses.dataclass
|
||||
class PoseLandmarkerResult:
|
||||
"""The pose landmarks detection result from PoseLandmarker, where each vector element represents a single pose detected in the image.
|
||||
|
||||
Attributes:
|
||||
pose_landmarks: Detected pose landmarks in normalized image coordinates.
|
||||
pose_world_landmarks: Detected pose landmarks in world coordinates.
|
||||
pose_auxiliary_landmarks: Detected auxiliary landmarks, used for deriving
|
||||
ROI for next frame.
|
||||
segmentation_masks: Segmentation masks for pose.
|
||||
"""
|
||||
|
||||
pose_landmarks: List[List[landmark_module.NormalizedLandmark]]
|
||||
pose_world_landmarks: List[List[landmark_module.Landmark]]
|
||||
pose_auxiliary_landmarks: List[List[landmark_module.NormalizedLandmark]]
|
||||
segmentation_masks: Optional[List[image_module.Image]] = None
|
||||
|
||||
|
||||
def _build_landmarker_result(
|
||||
output_packets: Mapping[str, packet_module.Packet]
|
||||
) -> PoseLandmarkerResult:
|
||||
"""Constructs a `PoseLandmarkerResult` from output packets."""
|
||||
pose_landmarker_result = PoseLandmarkerResult([], [], [], [])
|
||||
|
||||
if _SEGMENTATION_MASK_STREAM_NAME in output_packets:
|
||||
pose_landmarker_result.segmentation_masks = packet_getter.get_image_list(
|
||||
output_packets[_SEGMENTATION_MASK_STREAM_NAME]
|
||||
)
|
||||
|
||||
pose_landmarks_proto_list = packet_getter.get_proto_list(
|
||||
output_packets[_NORM_LANDMARKS_STREAM_NAME]
|
||||
)
|
||||
pose_world_landmarks_proto_list = packet_getter.get_proto_list(
|
||||
output_packets[_POSE_WORLD_LANDMARKS_STREAM_NAME]
|
||||
)
|
||||
pose_auxiliary_landmarks_proto_list = packet_getter.get_proto_list(
|
||||
output_packets[_POSE_AUXILIARY_LANDMARKS_STREAM_NAME]
|
||||
)
|
||||
|
||||
for proto in pose_landmarks_proto_list:
|
||||
pose_landmarks = landmark_pb2.NormalizedLandmarkList()
|
||||
pose_landmarks.MergeFrom(proto)
|
||||
pose_landmarks_list = []
|
||||
for pose_landmark in pose_landmarks.landmark:
|
||||
pose_landmarks_list.append(
|
||||
landmark_module.NormalizedLandmark.create_from_pb2(pose_landmark)
|
||||
)
|
||||
pose_landmarker_result.pose_landmarks.append(pose_landmarks_list)
|
||||
|
||||
for proto in pose_world_landmarks_proto_list:
|
||||
pose_world_landmarks = landmark_pb2.LandmarkList()
|
||||
pose_world_landmarks.MergeFrom(proto)
|
||||
pose_world_landmarks_list = []
|
||||
for pose_world_landmark in pose_world_landmarks.landmark:
|
||||
pose_world_landmarks_list.append(
|
||||
landmark_module.Landmark.create_from_pb2(pose_world_landmark)
|
||||
)
|
||||
pose_landmarker_result.pose_world_landmarks.append(
|
||||
pose_world_landmarks_list
|
||||
)
|
||||
|
||||
for proto in pose_auxiliary_landmarks_proto_list:
|
||||
pose_auxiliary_landmarks = landmark_pb2.NormalizedLandmarkList()
|
||||
pose_auxiliary_landmarks.MergeFrom(proto)
|
||||
pose_auxiliary_landmarks_list = []
|
||||
for pose_auxiliary_landmark in pose_auxiliary_landmarks.landmark:
|
||||
pose_auxiliary_landmarks_list.append(
|
||||
landmark_module.NormalizedLandmark.create_from_pb2(pose_auxiliary_landmark)
|
||||
)
|
||||
pose_landmarker_result.pose_auxiliary_landmarks.append(
|
||||
pose_auxiliary_landmarks_list
|
||||
)
|
||||
return pose_landmarker_result
|
||||
|
||||
|
||||
@dataclasses.dataclass
|
||||
class PoseLandmarkerOptions:
|
||||
"""Options for the pose landmarker task.
|
||||
|
||||
Attributes:
|
||||
base_options: Base options for the pose landmarker task.
|
||||
running_mode: The running mode of the task. Default to the image mode.
|
||||
HandLandmarker has three running modes: 1) The image mode for detecting
|
||||
pose landmarks on single image inputs. 2) The video mode for detecting
|
||||
pose landmarks on the decoded frames of a video. 3) The live stream mode
|
||||
for detecting pose landmarks on the live stream of input data, such as
|
||||
from camera. In this mode, the "result_callback" below must be specified
|
||||
to receive the detection results asynchronously.
|
||||
num_poses: The maximum number of poses can be detected by the PoseLandmarker.
|
||||
min_pose_detection_confidence: The minimum confidence score for the pose
|
||||
detection to be considered successful.
|
||||
min_pose_presence_confidence: The minimum confidence score of pose presence
|
||||
score in the pose landmark detection.
|
||||
min_tracking_confidence: The minimum confidence score for the pose tracking
|
||||
to be considered successful.
|
||||
result_callback: The user-defined result callback for processing live stream
|
||||
data. The result callback should only be specified when the running mode
|
||||
is set to the live stream mode.
|
||||
"""
|
||||
|
||||
base_options: _BaseOptions
|
||||
running_mode: _RunningMode = _RunningMode.IMAGE
|
||||
num_poses: int = 1
|
||||
min_pose_detection_confidence: float = 0.5
|
||||
min_pose_presence_confidence: float = 0.5
|
||||
min_tracking_confidence: float = 0.5
|
||||
output_segmentation_masks: bool = False
|
||||
result_callback: Optional[
|
||||
Callable[[PoseLandmarkerResult, image_module.Image, int], None]
|
||||
] = None
|
||||
|
||||
@doc_controls.do_not_generate_docs
|
||||
def to_pb2(self) -> _PoseLandmarkerGraphOptionsProto:
|
||||
"""Generates an PoseLandmarkerGraphOptions protobuf object."""
|
||||
base_options_proto = self.base_options.to_pb2()
|
||||
base_options_proto.use_stream_mode = (
|
||||
False if self.running_mode == _RunningMode.IMAGE else True
|
||||
)
|
||||
|
||||
# Initialize the pose landmarker options from base options.
|
||||
pose_landmarker_options_proto = _PoseLandmarkerGraphOptionsProto(
|
||||
base_options=base_options_proto
|
||||
)
|
||||
pose_landmarker_options_proto.min_tracking_confidence = (
|
||||
self.min_tracking_confidence
|
||||
)
|
||||
pose_landmarker_options_proto.pose_detector_graph_options.num_poses = (
|
||||
self.num_poses
|
||||
)
|
||||
pose_landmarker_options_proto.pose_detector_graph_options.min_detection_confidence = (
|
||||
self.min_pose_detection_confidence
|
||||
)
|
||||
pose_landmarker_options_proto.pose_landmarks_detector_graph_options.min_detection_confidence = (
|
||||
self.min_pose_presence_confidence
|
||||
)
|
||||
return pose_landmarker_options_proto
|
||||
|
||||
|
||||
class PoseLandmarker(base_vision_task_api.BaseVisionTaskApi):
|
||||
"""Class that performs pose landmarks detection on images."""
|
||||
|
||||
@classmethod
|
||||
def create_from_model_path(cls, model_path: str) -> 'PoseLandmarker':
|
||||
"""Creates an `PoseLandmarker` object from a TensorFlow Lite model and the default `PoseLandmarkerOptions`.
|
||||
|
||||
Note that the created `PoseLandmarker` instance is in image mode, for
|
||||
detecting pose landmarks on single image inputs.
|
||||
|
||||
Args:
|
||||
model_path: Path to the model.
|
||||
|
||||
Returns:
|
||||
`PoseLandmarker` object that's created from the model file and the
|
||||
default `PoseLandmarkerOptions`.
|
||||
|
||||
Raises:
|
||||
ValueError: If failed to create `PoseLandmarker` 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 = PoseLandmarkerOptions(
|
||||
base_options=base_options, running_mode=_RunningMode.IMAGE
|
||||
)
|
||||
return cls.create_from_options(options)
|
||||
|
||||
@classmethod
|
||||
def create_from_options(
|
||||
cls, options: PoseLandmarkerOptions
|
||||
) -> 'PoseLandmarker':
|
||||
"""Creates the `PoseLandmarker` object from pose landmarker options.
|
||||
|
||||
Args:
|
||||
options: Options for the pose landmarker task.
|
||||
|
||||
Returns:
|
||||
`PoseLandmarker` object that's created from `options`.
|
||||
|
||||
Raises:
|
||||
ValueError: If failed to create `PoseLandmarker` object from
|
||||
`PoseLandmarkerOptions` such as missing the model.
|
||||
RuntimeError: If other types of error occurred.
|
||||
"""
|
||||
|
||||
def packets_callback(output_packets: Mapping[str, packet_module.Packet]):
|
||||
if output_packets[_IMAGE_OUT_STREAM_NAME].is_empty():
|
||||
return
|
||||
|
||||
image = packet_getter.get_image(output_packets[_IMAGE_OUT_STREAM_NAME])
|
||||
|
||||
if output_packets[_NORM_LANDMARKS_STREAM_NAME].is_empty():
|
||||
empty_packet = output_packets[_NORM_LANDMARKS_STREAM_NAME]
|
||||
options.result_callback(
|
||||
PoseLandmarkerResult([], [], []),
|
||||
image,
|
||||
empty_packet.timestamp.value // _MICRO_SECONDS_PER_MILLISECOND,
|
||||
)
|
||||
return
|
||||
|
||||
pose_landmarker_result = _build_landmarker_result(output_packets)
|
||||
timestamp = output_packets[_NORM_LANDMARKS_STREAM_NAME].timestamp
|
||||
options.result_callback(
|
||||
pose_landmarker_result,
|
||||
image,
|
||||
timestamp.value // _MICRO_SECONDS_PER_MILLISECOND,
|
||||
)
|
||||
|
||||
output_streams = [
|
||||
':'.join([_SEGMENTATION_MASK_TAG, _SEGMENTATION_MASK_STREAM_NAME]),
|
||||
':'.join([_NORM_LANDMARKS_TAG, _NORM_LANDMARKS_STREAM_NAME]),
|
||||
':'.join([
|
||||
_POSE_WORLD_LANDMARKS_TAG, _POSE_WORLD_LANDMARKS_STREAM_NAME
|
||||
]),
|
||||
':'.join([
|
||||
_POSE_AUXILIARY_LANDMARKS_TAG,
|
||||
_POSE_AUXILIARY_LANDMARKS_STREAM_NAME
|
||||
]),
|
||||
':'.join([_IMAGE_TAG, _IMAGE_OUT_STREAM_NAME]),
|
||||
]
|
||||
|
||||
if options.output_segmentation_masks:
|
||||
output_streams.append(
|
||||
':'.join([_SEGMENTATION_MASK_TAG, _SEGMENTATION_MASK_STREAM_NAME])
|
||||
)
|
||||
|
||||
task_info = _TaskInfo(
|
||||
task_graph=_TASK_GRAPH_NAME,
|
||||
input_streams=[
|
||||
':'.join([_IMAGE_TAG, _IMAGE_IN_STREAM_NAME]),
|
||||
':'.join([_NORM_RECT_TAG, _NORM_RECT_STREAM_NAME]),
|
||||
],
|
||||
output_streams=output_streams,
|
||||
task_options=options,
|
||||
)
|
||||
return cls(
|
||||
task_info.generate_graph_config(
|
||||
enable_flow_limiting=options.running_mode
|
||||
== _RunningMode.LIVE_STREAM
|
||||
),
|
||||
options.running_mode,
|
||||
packets_callback if options.result_callback else None,
|
||||
)
|
||||
|
||||
def detect(
|
||||
self,
|
||||
image: image_module.Image,
|
||||
image_processing_options: Optional[_ImageProcessingOptions] = None,
|
||||
) -> PoseLandmarkerResult:
|
||||
"""Performs pose landmarks detection on the given image.
|
||||
|
||||
Only use this method when the PoseLandmarker is created with the image
|
||||
running mode.
|
||||
|
||||
Args:
|
||||
image: MediaPipe Image.
|
||||
image_processing_options: Options for image processing.
|
||||
|
||||
Returns:
|
||||
The pose landmarker detection results.
|
||||
|
||||
Raises:
|
||||
ValueError: If any of the input arguments is invalid.
|
||||
RuntimeError: If pose landmarker detection failed to run.
|
||||
"""
|
||||
normalized_rect = self.convert_to_normalized_rect(
|
||||
image_processing_options, image, roi_allowed=False
|
||||
)
|
||||
output_packets = self._process_image_data({
|
||||
_IMAGE_IN_STREAM_NAME: packet_creator.create_image(image),
|
||||
_NORM_RECT_STREAM_NAME: packet_creator.create_proto(
|
||||
normalized_rect.to_pb2()
|
||||
),
|
||||
})
|
||||
|
||||
if output_packets[_NORM_LANDMARKS_STREAM_NAME].is_empty():
|
||||
return PoseLandmarkerResult([], [], [])
|
||||
|
||||
return _build_landmarker_result(output_packets)
|
||||
|
||||
def detect_for_video(
|
||||
self,
|
||||
image: image_module.Image,
|
||||
timestamp_ms: int,
|
||||
image_processing_options: Optional[_ImageProcessingOptions] = None,
|
||||
) -> PoseLandmarkerResult:
|
||||
"""Performs pose landmarks detection on the provided video frame.
|
||||
|
||||
Only use this method when the PoseLandmarker is created with the video
|
||||
running mode.
|
||||
|
||||
Only use this method when the PoseLandmarker is created with the video
|
||||
running mode. It's required to provide the video frame's timestamp (in
|
||||
milliseconds) along with the video frame. The input timestamps should be
|
||||
monotonically increasing for adjacent calls of this method.
|
||||
|
||||
Args:
|
||||
image: MediaPipe Image.
|
||||
timestamp_ms: The timestamp of the input video frame in milliseconds.
|
||||
image_processing_options: Options for image processing.
|
||||
|
||||
Returns:
|
||||
The pose landmarks detection results.
|
||||
|
||||
Raises:
|
||||
ValueError: If any of the input arguments is invalid.
|
||||
RuntimeError: If pose landmarker detection failed to run.
|
||||
"""
|
||||
normalized_rect = self.convert_to_normalized_rect(
|
||||
image_processing_options, image, roi_allowed=False
|
||||
)
|
||||
output_packets = self._process_video_data({
|
||||
_IMAGE_IN_STREAM_NAME: packet_creator.create_image(image).at(
|
||||
timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND
|
||||
),
|
||||
_NORM_RECT_STREAM_NAME: packet_creator.create_proto(
|
||||
normalized_rect.to_pb2()
|
||||
).at(timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND),
|
||||
})
|
||||
|
||||
if output_packets[_NORM_LANDMARKS_STREAM_NAME].is_empty():
|
||||
return PoseLandmarkerResult([], [], [])
|
||||
|
||||
return _build_landmarker_result(output_packets)
|
||||
|
||||
def detect_async(
|
||||
self,
|
||||
image: image_module.Image,
|
||||
timestamp_ms: int,
|
||||
image_processing_options: Optional[_ImageProcessingOptions] = None,
|
||||
) -> None:
|
||||
"""Sends live image data to perform pose landmarks detection.
|
||||
|
||||
The results will be available via the "result_callback" provided in the
|
||||
PoseLandmarkerOptions. Only use this method when the PoseLandmarker is
|
||||
created with the live stream running mode.
|
||||
|
||||
Only use this method when the PoseLandmarker is created with the live
|
||||
stream running mode. The input timestamps should be monotonically increasing
|
||||
for adjacent calls of this method. This method will return immediately after
|
||||
the input image is accepted. The results will be available via the
|
||||
`result_callback` provided in the `PoseLandmarkerOptions`. The
|
||||
`detect_async` method is designed to process live stream data such as
|
||||
camera input. To lower the overall latency, pose landmarker may drop the
|
||||
input images if needed. In other words, it's not guaranteed to have output
|
||||
per input image.
|
||||
|
||||
The `result_callback` provides:
|
||||
- The pose landmarks detection results.
|
||||
- The input image that the pose landmarker runs on.
|
||||
- The input timestamp in milliseconds.
|
||||
|
||||
Args:
|
||||
image: MediaPipe Image.
|
||||
timestamp_ms: The timestamp of the input image in milliseconds.
|
||||
image_processing_options: Options for image processing.
|
||||
|
||||
Raises:
|
||||
ValueError: If the current input timestamp is smaller than what the
|
||||
pose landmarker has already processed.
|
||||
"""
|
||||
normalized_rect = self.convert_to_normalized_rect(
|
||||
image_processing_options, image, roi_allowed=False
|
||||
)
|
||||
self._send_live_stream_data({
|
||||
_IMAGE_IN_STREAM_NAME: packet_creator.create_image(image).at(
|
||||
timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND
|
||||
),
|
||||
_NORM_RECT_STREAM_NAME: packet_creator.create_proto(
|
||||
normalized_rect.to_pb2()
|
||||
).at(timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND),
|
||||
})
|
||||
Loading…
Reference in New Issue
Block a user