Merge pull request #4303 from kinaryml:pose-landmarker-python

PiperOrigin-RevId: 527948047

mediapipe/python/BUILD

@@ -99,6 +99,7 @@ cc_library(
         "//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/pose_landmarker:pose_landmarker_graph",
     ] + select({
         # TODO: Build text_classifier_graph and text_embedder_graph on Windows.
         "//mediapipe:windows": [],

mediapipe/tasks/python/test/vision/BUILD

@@ -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/python:_framework_bindings",
+        "//mediapipe/tasks/cc/components/containers/proto:landmarks_detection_result_py_pb2",
+        "//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",
+    ],
+)

mediapipe/tasks/python/test/vision/face_landmarker_test.py

@@ -51,24 +51,27 @@ _PORTRAIT_IMAGE = 'portrait.jpg'
 _CAT_IMAGE = 'cat.jpg'
 _PORTRAIT_EXPECTED_FACE_LANDMARKS = 'portrait_expected_face_landmarks.pbtxt'
 _PORTRAIT_EXPECTED_BLENDSHAPES = 'portrait_expected_blendshapes.pbtxt'
-_LANDMARKS_DIFF_MARGIN = 0.03
+_LANDMARKS_MARGIN = 0.03
-_BLENDSHAPES_DIFF_MARGIN = 0.13
+_BLENDSHAPES_MARGIN = 0.13
-_FACIAL_TRANSFORMATION_MATRIX_DIFF_MARGIN = 0.02
+_FACIAL_TRANSFORMATION_MATRIX_MARGIN = 0.02
 
 
 def _get_expected_face_landmarks(file_path: str):
   proto_file_path = test_utils.get_test_data_path(file_path)
+  face_landmarks_results = []
   with open(proto_file_path, 'rb') as f:
     proto = landmark_pb2.NormalizedLandmarkList()
     text_format.Parse(f.read(), proto)
     face_landmarks = []
     for landmark in proto.landmark:
       face_landmarks.append(_NormalizedLandmark.create_from_pb2(landmark))
-  return face_landmarks
+  face_landmarks_results.append(face_landmarks)
+  return face_landmarks_results
 
 
 def _get_expected_face_blendshapes(file_path: str):
   proto_file_path = test_utils.get_test_data_path(file_path)
+  face_blendshapes_results = []
   with open(proto_file_path, 'rb') as f:
     proto = classification_pb2.ClassificationList()
     text_format.Parse(f.read(), proto)
@@ -84,7 +87,8 @@ def _get_expected_face_blendshapes(file_path: str):
               category_name=face_blendshapes.label,
           )
       )
-  return face_blendshapes_categories
+  face_blendshapes_results.append(face_blendshapes_categories)
+  return face_blendshapes_results
 
 
 def _get_expected_facial_transformation_matrixes():
@@ -119,13 +123,14 @@ class FaceLandmarkerTest(parameterized.TestCase):
     # Expects to have the same number of faces detected.
     self.assertLen(actual_landmarks, len(expected_landmarks))
 
-    for i, elem in enumerate(actual_landmarks):
+    for i, _ in enumerate(actual_landmarks):
-      self.assertAlmostEqual(
+      for j, elem in enumerate(actual_landmarks[i]):
-          elem.x, expected_landmarks[i].x, delta=_LANDMARKS_DIFF_MARGIN
+        self.assertAlmostEqual(
-      )
+            elem.x, expected_landmarks[i][j].x, delta=_LANDMARKS_MARGIN
-      self.assertAlmostEqual(
+        )
-          elem.y, expected_landmarks[i].y, delta=_LANDMARKS_DIFF_MARGIN
+        self.assertAlmostEqual(
-      )
+            elem.y, expected_landmarks[i][j].y, delta=_LANDMARKS_MARGIN
+        )
 
   def _expect_blendshapes_correct(
       self, actual_blendshapes, expected_blendshapes
@@ -133,13 +138,14 @@ class FaceLandmarkerTest(parameterized.TestCase):
     # Expects to have the same number of blendshapes.
     self.assertLen(actual_blendshapes, len(expected_blendshapes))
 
-    for i, elem in enumerate(actual_blendshapes):
+    for i, _ in enumerate(actual_blendshapes):
-      self.assertEqual(elem.index, expected_blendshapes[i].index)
+      for j, elem in enumerate(actual_blendshapes[i]):
-      self.assertAlmostEqual(
+        self.assertEqual(elem.index, expected_blendshapes[i][j].index)
-          elem.score,
+        self.assertAlmostEqual(
-          expected_blendshapes[i].score,
+            elem.score,
-          delta=_BLENDSHAPES_DIFF_MARGIN,
+            expected_blendshapes[i][j].score,
-      )
+            delta=_BLENDSHAPES_MARGIN,
+        )
 
   def _expect_facial_transformation_matrixes_correct(
       self, actual_matrix_list, expected_matrix_list
@@ -152,7 +158,7 @@ class FaceLandmarkerTest(parameterized.TestCase):
       self.assertSequenceAlmostEqual(
           elem.flatten(),
           expected_matrix_list[i].flatten(),
-          delta=_FACIAL_TRANSFORMATION_MATRIX_DIFF_MARGIN,
+          delta=_FACIAL_TRANSFORMATION_MATRIX_MARGIN,
       )
 
   def test_create_from_file_succeeds_with_valid_model_path(self):
@@ -236,11 +242,11 @@ class FaceLandmarkerTest(parameterized.TestCase):
     # Comparing results.
     if expected_face_landmarks is not None:
       self._expect_landmarks_correct(
-          detection_result.face_landmarks[0], expected_face_landmarks
+          detection_result.face_landmarks, expected_face_landmarks
       )
     if expected_face_blendshapes is not None:
       self._expect_blendshapes_correct(
-          detection_result.face_blendshapes[0], expected_face_blendshapes
+          detection_result.face_blendshapes, expected_face_blendshapes
       )
     if expected_facial_transformation_matrixes is not None:
       self._expect_facial_transformation_matrixes_correct(
@@ -302,11 +308,11 @@ class FaceLandmarkerTest(parameterized.TestCase):
       # Comparing results.
       if expected_face_landmarks is not None:
         self._expect_landmarks_correct(
-            detection_result.face_landmarks[0], expected_face_landmarks
+            detection_result.face_landmarks, expected_face_landmarks
         )
       if expected_face_blendshapes is not None:
         self._expect_blendshapes_correct(
-            detection_result.face_blendshapes[0], expected_face_blendshapes
+            detection_result.face_blendshapes, expected_face_blendshapes
         )
       if expected_facial_transformation_matrixes is not None:
         self._expect_facial_transformation_matrixes_correct(
@@ -446,11 +452,11 @@ class FaceLandmarkerTest(parameterized.TestCase):
         # Comparing results.
         if expected_face_landmarks is not None:
           self._expect_landmarks_correct(
-              detection_result.face_landmarks[0], expected_face_landmarks
+              detection_result.face_landmarks, expected_face_landmarks
           )
         if expected_face_blendshapes is not None:
           self._expect_blendshapes_correct(
-              detection_result.face_blendshapes[0], expected_face_blendshapes
+              detection_result.face_blendshapes, expected_face_blendshapes
           )
         if expected_facial_transformation_matrixes is not None:
           self._expect_facial_transformation_matrixes_correct(
@@ -523,11 +529,11 @@ class FaceLandmarkerTest(parameterized.TestCase):
       # Comparing results.
       if expected_face_landmarks is not None:
         self._expect_landmarks_correct(
-            result.face_landmarks[0], expected_face_landmarks
+            result.face_landmarks, expected_face_landmarks
         )
       if expected_face_blendshapes is not None:
         self._expect_blendshapes_correct(
-            result.face_blendshapes[0], expected_face_blendshapes
+            result.face_blendshapes, expected_face_blendshapes
         )
       if expected_facial_transformation_matrixes is not None:
         self._expect_facial_transformation_matrixes_correct(

mediapipe/tasks/python/test/vision/hand_landmarker_test.py

@@ -32,12 +32,14 @@ from mediapipe.tasks.python.vision import hand_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
 
-_LandmarksDetectionResultProto = landmarks_detection_result_pb2.LandmarksDetectionResult
+_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
+_LandmarksDetectionResult = (
+    landmark_detection_result_module.LandmarksDetectionResult)
 _Image = image_module.Image
 _HandLandmarker = hand_landmarker.HandLandmarker
 _HandLandmarkerOptions = hand_landmarker.HandLandmarkerOptions
@@ -54,7 +56,7 @@ _POINTING_UP_IMAGE = 'pointing_up.jpg'
 _POINTING_UP_LANDMARKS = 'pointing_up_landmarks.pbtxt'
 _POINTING_UP_ROTATED_IMAGE = 'pointing_up_rotated.jpg'
 _POINTING_UP_ROTATED_LANDMARKS = 'pointing_up_rotated_landmarks.pbtxt'
-_LANDMARKS_ERROR_TOLERANCE = 0.03
+_LANDMARKS_MARGIN = 0.03
 _HANDEDNESS_MARGIN = 0.05
 
 
@@ -89,39 +91,43 @@ class HandLandmarkerTest(parameterized.TestCase):
     self.model_path = test_utils.get_test_data_path(
         _HAND_LANDMARKER_BUNDLE_ASSET_FILE)
 
-  def _assert_actual_result_approximately_matches_expected_result(
+  def _expect_hand_landmarks_correct(
-      self, actual_result: _HandLandmarkerResult,
+      self, actual_landmarks, expected_landmarks, margin
-      expected_result: _HandLandmarkerResult):
+  ):
     # Expects to have the same number of hands detected.
-    self.assertLen(actual_result.hand_landmarks,
+    self.assertLen(actual_landmarks, len(expected_landmarks))
-                   len(expected_result.hand_landmarks))
+
-    self.assertLen(actual_result.hand_world_landmarks,
+    for i, _ in enumerate(actual_landmarks):
-                   len(expected_result.hand_world_landmarks))
+      for j, elem in enumerate(actual_landmarks[i]):
-    self.assertLen(actual_result.handedness, len(expected_result.handedness))
+        self.assertAlmostEqual(elem.x, expected_landmarks[i][j].x, delta=margin)
-    # Actual landmarks match expected landmarks.
+        self.assertAlmostEqual(elem.y, expected_landmarks[i][j].y, delta=margin)
-    self.assertLen(actual_result.hand_landmarks[0],
+
-                   len(expected_result.hand_landmarks[0]))
+  def _expect_handedness_correct(
-    actual_landmarks = actual_result.hand_landmarks[0]
+      self, actual_handedness, expected_handedness, margin
-    expected_landmarks = expected_result.hand_landmarks[0]
+  ):
-    for i, rename_me in enumerate(actual_landmarks):
+    # Actual top handedness matches expected top handedness.
-      self.assertAlmostEqual(
+    actual_top_handedness = actual_handedness[0][0]
-          rename_me.x,
+    expected_top_handedness = expected_handedness[0][0]
-          expected_landmarks[i].x,
-          delta=_LANDMARKS_ERROR_TOLERANCE)
-      self.assertAlmostEqual(
-          rename_me.y,
-          expected_landmarks[i].y,
-          delta=_LANDMARKS_ERROR_TOLERANCE)
-    # Actual handedness matches expected handedness.
-    actual_top_handedness = actual_result.handedness[0][0]
-    expected_top_handedness = expected_result.handedness[0][0]
     self.assertEqual(actual_top_handedness.index, expected_top_handedness.index)
     self.assertEqual(actual_top_handedness.category_name,
                      expected_top_handedness.category_name)
     self.assertAlmostEqual(
-        actual_top_handedness.score,
+        actual_top_handedness.score, expected_top_handedness.score, delta=margin
-        expected_top_handedness.score,
+    )
-        delta=_HANDEDNESS_MARGIN)
+
+  def _expect_hand_landmarker_results_correct(
+      self,
+      actual_result: _HandLandmarkerResult,
+      expected_result: _HandLandmarkerResult,
+  ):
+    self._expect_hand_landmarks_correct(
+        actual_result.hand_landmarks,
+        expected_result.hand_landmarks,
+        _LANDMARKS_MARGIN,
+    )
+    self._expect_handedness_correct(
+        actual_result.handedness, expected_result.handedness, _HANDEDNESS_MARGIN
+    )
 
   def test_create_from_file_succeeds_with_valid_model_path(self):
     # Creates with default option and valid model file successfully.
@@ -175,8 +181,9 @@ class HandLandmarkerTest(parameterized.TestCase):
     # Performs hand landmarks detection on the input.
     detection_result = landmarker.detect(self.test_image)
     # Comparing results.
-    self._assert_actual_result_approximately_matches_expected_result(
+    self._expect_hand_landmarker_results_correct(
-        detection_result, expected_detection_result)
+        detection_result, expected_detection_result
+    )
     # Closes the hand landmarker explicitly when the hand landmarker is not used
     # in a context.
     landmarker.close()
@@ -203,8 +210,9 @@ class HandLandmarkerTest(parameterized.TestCase):
       # Performs hand landmarks detection on the input.
       detection_result = landmarker.detect(self.test_image)
       # Comparing results.
-      self._assert_actual_result_approximately_matches_expected_result(
+      self._expect_hand_landmarker_results_correct(
-          detection_result, expected_detection_result)
+          detection_result, expected_detection_result
+      )
 
   def test_detect_succeeds_with_num_hands(self):
     # Creates hand landmarker.
@@ -234,8 +242,9 @@ class HandLandmarkerTest(parameterized.TestCase):
       expected_detection_result = _get_expected_hand_landmarker_result(
           _POINTING_UP_ROTATED_LANDMARKS)
       # Comparing results.
-      self._assert_actual_result_approximately_matches_expected_result(
+      self._expect_hand_landmarker_results_correct(
-          detection_result, expected_detection_result)
+          detection_result, expected_detection_result
+      )
 
   def test_detect_fails_with_region_of_interest(self):
     # Creates hand landmarker.
@@ -350,9 +359,9 @@ class HandLandmarkerTest(parameterized.TestCase):
       for timestamp in range(0, 300, 30):
         result = landmarker.detect_for_video(test_image, timestamp,
                                              image_processing_options)
-        if result.hand_landmarks and result.hand_world_landmarks and result.handedness:
+        if (result.hand_landmarks and result.hand_world_landmarks and
-          self._assert_actual_result_approximately_matches_expected_result(
+            result.handedness):
-              result, expected_result)
+          self._expect_hand_landmarker_results_correct(result, expected_result)
         else:
           self.assertEqual(result, expected_result)
 
@@ -405,9 +414,9 @@ class HandLandmarkerTest(parameterized.TestCase):
 
     def check_result(result: _HandLandmarkerResult, output_image: _Image,
                      timestamp_ms: int):
-      if result.hand_landmarks and result.hand_world_landmarks and result.handedness:
+      if (result.hand_landmarks and result.hand_world_landmarks and
-        self._assert_actual_result_approximately_matches_expected_result(
+          result.handedness):
-            result, expected_result)
+        self._expect_hand_landmarker_results_correct(result, expected_result)
       else:
         self.assertEqual(result, expected_result)
       self.assertTrue(

mediapipe/tasks/python/test/vision/pose_landmarker_test.py

@@ -0,0 +1,520 @@
+# 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 pose landmarker."""
+
+import enum
+from typing import List
+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_MARGIN = 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_landmarks_correct(
+      self, actual_landmarks, expected_landmarks, margin
+  ):
+    # Expects to have the same number of poses detected.
+    self.assertLen(actual_landmarks, len(expected_landmarks))
+
+    for i, _ in enumerate(actual_landmarks):
+      for j, elem in enumerate(actual_landmarks[i]):
+        self.assertAlmostEqual(elem.x, expected_landmarks[i][j].x, delta=margin)
+        self.assertAlmostEqual(elem.y, expected_landmarks[i][j].y, delta=margin)
+
+  def _expect_pose_landmarker_results_correct(
+      self,
+      actual_result: PoseLandmarkerResult,
+      expected_result: PoseLandmarkerResult,
+      output_segmentation_masks: bool,
+      margin: float,
+  ):
+    self._expect_pose_landmarks_correct(
+        actual_result.pose_landmarks, expected_result.pose_landmarks, margin
+    )
+    if output_segmentation_masks:
+      self.assertIsInstance(actual_result.segmentation_masks, List)
+      for _, mask in enumerate(actual_result.segmentation_masks):
+        self.assertIsInstance(mask, _Image)
+    else:
+      self.assertIsNone(actual_result.segmentation_masks)
+
+  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,
+          False,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+      (
+          ModelFileType.FILE_CONTENT,
+          False,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+      (
+          ModelFileType.FILE_NAME,
+          True,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+      (
+          ModelFileType.FILE_CONTENT,
+          True,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+  )
+  def test_detect(
+      self,
+      model_file_type,
+      output_segmentation_masks,
+      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,
+        output_segmentation_masks=output_segmentation_masks,
+    )
+    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,
+        output_segmentation_masks,
+        _LANDMARKS_MARGIN,
+    )
+    # Closes the pose landmarker explicitly when the pose landmarker is not used
+    # in a context.
+    landmarker.close()
+
+  @parameterized.parameters(
+      (
+          ModelFileType.FILE_NAME,
+          False,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+      (
+          ModelFileType.FILE_CONTENT,
+          False,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+      (
+          ModelFileType.FILE_NAME,
+          True,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+      (
+          ModelFileType.FILE_CONTENT,
+          True,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+  )
+  def test_detect_in_context(
+      self,
+      model_file_type,
+      output_segmentation_masks,
+      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,
+        output_segmentation_masks=output_segmentation_masks,
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      # 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,
+          output_segmentation_masks,
+          _LANDMARKS_MARGIN,
+      )
+
+  def test_detect_fails_with_region_of_interest(self):
+    # Creates pose landmarker.
+    base_options = _BaseOptions(model_asset_path=self.model_path)
+    options = _PoseLandmarkerOptions(base_options=base_options)
+    with self.assertRaisesRegex(
+        ValueError, "This task doesn't support region-of-interest."
+    ):
+      with _PoseLandmarker.create_from_options(options) as landmarker:
+        # Set the `region_of_interest` parameter using `ImageProcessingOptions`.
+        image_processing_options = _ImageProcessingOptions(
+            region_of_interest=_Rect(0, 0, 1, 1)
+        )
+        # Attempt to perform pose landmarks detection on the cropped input.
+        landmarker.detect(self.test_image, image_processing_options)
+
+  def test_empty_detection_outputs(self):
+    # Creates pose landmarker.
+    base_options = _BaseOptions(model_asset_path=self.model_path)
+    options = _PoseLandmarkerOptions(base_options=base_options)
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      # Load an image with no poses.
+      test_image = _Image.create_from_file(
+          test_utils.get_test_data_path(_BURGER_IMAGE)
+      )
+      # Performs pose landmarks detection on the input.
+      detection_result = landmarker.detect(test_image)
+      # Comparing results.
+      self.assertEmpty(detection_result.pose_landmarks)
+      self.assertEmpty(detection_result.pose_world_landmarks)
+      self.assertEmpty(detection_result.pose_auxiliary_landmarks)
+
+  def test_missing_result_callback(self):
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+    )
+    with self.assertRaisesRegex(
+        ValueError, r'result callback must be provided'
+    ):
+      with _PoseLandmarker.create_from_options(options) as unused_landmarker:
+        pass
+
+  @parameterized.parameters((_RUNNING_MODE.IMAGE), (_RUNNING_MODE.VIDEO))
+  def test_illegal_result_callback(self, running_mode):
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=running_mode,
+        result_callback=mock.MagicMock(),
+    )
+    with self.assertRaisesRegex(
+        ValueError, r'result callback should not be provided'
+    ):
+      with _PoseLandmarker.create_from_options(options) as unused_landmarker:
+        pass
+
+  def test_calling_detect_for_video_in_image_mode(self):
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.IMAGE,
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the video mode'
+      ):
+        landmarker.detect_for_video(self.test_image, 0)
+
+  def test_calling_detect_async_in_image_mode(self):
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.IMAGE,
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the live stream mode'
+      ):
+        landmarker.detect_async(self.test_image, 0)
+
+  def test_calling_detect_in_video_mode(self):
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO,
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the image mode'
+      ):
+        landmarker.detect(self.test_image)
+
+  def test_calling_detect_async_in_video_mode(self):
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO,
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the live stream mode'
+      ):
+        landmarker.detect_async(self.test_image, 0)
+
+  def test_detect_for_video_with_out_of_order_timestamp(self):
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO,
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      unused_result = landmarker.detect_for_video(self.test_image, 1)
+      with self.assertRaisesRegex(
+          ValueError, r'Input timestamp must be monotonically increasing'
+      ):
+        landmarker.detect_for_video(self.test_image, 0)
+
+  @parameterized.parameters(
+      (
+          _POSE_IMAGE,
+          0,
+          False,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+      (
+          _POSE_IMAGE,
+          0,
+          True,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+      (_BURGER_IMAGE, 0, False, PoseLandmarkerResult([], [], [])),
+  )
+  def test_detect_for_video(
+      self, image_path, rotation, output_segmentation_masks, expected_result
+  ):
+    test_image = _Image.create_from_file(
+        test_utils.get_test_data_path(image_path)
+    )
+    # Set rotation parameters using ImageProcessingOptions.
+    image_processing_options = _ImageProcessingOptions(
+        rotation_degrees=rotation
+    )
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        output_segmentation_masks=output_segmentation_masks,
+        running_mode=_RUNNING_MODE.VIDEO,
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      for timestamp in range(0, 300, 30):
+        result = landmarker.detect_for_video(
+            test_image, timestamp, image_processing_options
+        )
+        if result.pose_landmarks:
+          self._expect_pose_landmarker_results_correct(
+              result,
+              expected_result,
+              output_segmentation_masks,
+              _LANDMARKS_MARGIN,
+          )
+        else:
+          self.assertEqual(result, expected_result)
+
+  def test_calling_detect_in_live_stream_mode(self):
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=mock.MagicMock(),
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the image mode'
+      ):
+        landmarker.detect(self.test_image)
+
+  def test_calling_detect_for_video_in_live_stream_mode(self):
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=mock.MagicMock(),
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      with self.assertRaisesRegex(
+          ValueError, r'not initialized with the video mode'
+      ):
+        landmarker.detect_for_video(self.test_image, 0)
+
+  def test_detect_async_calls_with_illegal_timestamp(self):
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=mock.MagicMock(),
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      landmarker.detect_async(self.test_image, 100)
+      with self.assertRaisesRegex(
+          ValueError, r'Input timestamp must be monotonically increasing'
+      ):
+        landmarker.detect_async(self.test_image, 0)
+
+  @parameterized.parameters(
+      (
+          _POSE_IMAGE,
+          0,
+          False,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+      (
+          _POSE_IMAGE,
+          0,
+          True,
+          _get_expected_pose_landmarker_result(_POSE_LANDMARKS),
+      ),
+      (_BURGER_IMAGE, 0, False, PoseLandmarkerResult([], [], [])),
+  )
+  def test_detect_async_calls(
+      self, image_path, rotation, output_segmentation_masks, expected_result
+  ):
+    test_image = _Image.create_from_file(
+        test_utils.get_test_data_path(image_path)
+    )
+    # Set rotation parameters using ImageProcessingOptions.
+    image_processing_options = _ImageProcessingOptions(
+        rotation_degrees=rotation
+    )
+    observed_timestamp_ms = -1
+
+    def check_result(
+        result: PoseLandmarkerResult, output_image: _Image, timestamp_ms: int
+    ):
+      if result.pose_landmarks:
+        self._expect_pose_landmarker_results_correct(
+            result,
+            expected_result,
+            output_segmentation_masks,
+            _LANDMARKS_MARGIN,
+        )
+      else:
+        self.assertEqual(result, expected_result)
+      self.assertTrue(
+          np.array_equal(output_image.numpy_view(), test_image.numpy_view())
+      )
+      self.assertLess(observed_timestamp_ms, timestamp_ms)
+      self.observed_timestamp_ms = timestamp_ms
+
+    options = _PoseLandmarkerOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        output_segmentation_masks=output_segmentation_masks,
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=check_result,
+    )
+    with _PoseLandmarker.create_from_options(options) as landmarker:
+      for timestamp in range(0, 300, 30):
+        landmarker.detect_async(test_image, timestamp, image_processing_options)
+
+
+if __name__ == '__main__':
+  absltest.main()

mediapipe/tasks/python/vision/BUILD

@@ -180,6 +180,27 @@ 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 = [

mediapipe/tasks/python/vision/face_detector.py

@@ -71,8 +71,8 @@ class FaceDetectorOptions:
 
   base_options: _BaseOptions
   running_mode: _RunningMode = _RunningMode.IMAGE
-  min_detection_confidence: Optional[float] = None
+  min_detection_confidence: float = 0.5
-  min_suppression_threshold: Optional[float] = None
+  min_suppression_threshold: float = 0.3
   result_callback: Optional[
       Callable[
           [detections_module.DetectionResult, image_module.Image, int], None

mediapipe/tasks/python/vision/face_landmarker.py

@@ -2966,12 +2966,12 @@ class FaceLandmarkerOptions:
 
   base_options: _BaseOptions
   running_mode: _RunningMode = _RunningMode.IMAGE
-  num_faces: Optional[int] = 1
+  num_faces: int = 1
-  min_face_detection_confidence: Optional[float] = 0.5
+  min_face_detection_confidence: float = 0.5
-  min_face_presence_confidence: Optional[float] = 0.5
+  min_face_presence_confidence: float = 0.5
-  min_tracking_confidence: Optional[float] = 0.5
+  min_tracking_confidence: float = 0.5
-  output_face_blendshapes: Optional[bool] = False
+  output_face_blendshapes: bool = False
-  output_facial_transformation_matrixes: Optional[bool] = False
+  output_facial_transformation_matrixes: bool = False
   result_callback: Optional[
       Callable[[FaceLandmarkerResult, image_module.Image, int], None]
   ] = None

mediapipe/tasks/python/vision/gesture_recognizer.py

@@ -194,15 +194,15 @@ class GestureRecognizerOptions:
 
   base_options: _BaseOptions
   running_mode: _RunningMode = _RunningMode.IMAGE
-  num_hands: Optional[int] = 1
+  num_hands: int = 1
-  min_hand_detection_confidence: Optional[float] = 0.5
+  min_hand_detection_confidence: float = 0.5
-  min_hand_presence_confidence: Optional[float] = 0.5
+  min_hand_presence_confidence: float = 0.5
-  min_tracking_confidence: Optional[float] = 0.5
+  min_tracking_confidence: float = 0.5
-  canned_gesture_classifier_options: Optional[_ClassifierOptions] = (
+  canned_gesture_classifier_options: _ClassifierOptions = dataclasses.field(
-      dataclasses.field(default_factory=_ClassifierOptions)
+      default_factory=_ClassifierOptions
   )
-  custom_gesture_classifier_options: Optional[_ClassifierOptions] = (
+  custom_gesture_classifier_options: _ClassifierOptions = dataclasses.field(
-      dataclasses.field(default_factory=_ClassifierOptions)
+      default_factory=_ClassifierOptions
   )
   result_callback: Optional[
       Callable[[GestureRecognizerResult, image_module.Image, int], None]

mediapipe/tasks/python/vision/hand_landmarker.py

@@ -182,10 +182,10 @@ class HandLandmarkerOptions:
 
   base_options: _BaseOptions
   running_mode: _RunningMode = _RunningMode.IMAGE
-  num_hands: Optional[int] = 1
+  num_hands: int = 1
-  min_hand_detection_confidence: Optional[float] = 0.5
+  min_hand_detection_confidence: float = 0.5
-  min_hand_presence_confidence: Optional[float] = 0.5
+  min_hand_presence_confidence: float = 0.5
-  min_tracking_confidence: Optional[float] = 0.5
+  min_tracking_confidence: float = 0.5
   result_callback: Optional[
       Callable[[HandLandmarkerResult, image_module.Image, int], None]
   ] = None

mediapipe/tasks/python/vision/pose_landmarker.py

@@ -0,0 +1,431 @@
+# 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 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: Optional 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.
+      PoseLandmarker 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.
+    output_segmentation_masks: whether to output segmentation masks.
+    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 a `PoseLandmarker` object from a model bundle file 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([_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 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_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 landmarker 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),
+    })