MediaPipe Team a9b643e0f5 Project import generated by Copybara.

GitOrigin-RevId: ff83882955f1a1e2a043ff4e71278be9d7217bbe

2021-05-05 14:56:16 -04:00

18 KiB

Raw Blame History

layout	title	parent	nav_order
default	Holistic	Solutions	6

MediaPipe Holistic

{: .no_toc }

Table of contents

{: .text-delta } 1. TOC {:toc}

---

Overview

Live perception of simultaneous human pose, face landmarks, and hand tracking in real-time on mobile devices can enable various modern life applications: fitness and sport analysis, gesture control and sign language recognition, augmented reality try-on and effects. MediaPipe already offers fast and accurate, yet separate, solutions for these tasks. Combining them all in real-time into a semantically consistent end-to-end solution is a uniquely difficult problem requiring simultaneous inference of multiple, dependent neural networks.


Fig 1. Example of MediaPipe Holistic.

ML Pipeline

The MediaPipe Holistic pipeline integrates separate models for pose, face and hand components, each of which are optimized for their particular domain. However, because of their different specializations, the input to one component is not well-suited for the others. The pose estimation model, for example, takes a lower, fixed resolution video frame (256x256) as input. But if one were to crop the hand and face regions from that image to pass to their respective models, the image resolution would be too low for accurate articulation. Therefore, we designed MediaPipe Holistic as a multi-stage pipeline, which treats the different regions using a region appropriate image resolution.

First, we estimate the human pose (top of Fig 2) with BlazePose’s pose detector and subsequent landmark model. Then, using the inferred pose landmarks we derive three regions of interest (ROI) crops for each hand (2x) and the face, and employ a re-crop model to improve the ROI. We then crop the full-resolution input frame to these ROIs and apply task-specific face and hand models to estimate their corresponding landmarks. Finally, we merge all landmarks with those of the pose model to yield the full 540+ landmarks.


Fig 2. MediaPipe Holistic Pipeline Overview.

To streamline the identification of ROIs for face and hands, we utilize a tracking approach similar to the one we use for standalone face and hand pipelines. It assumes that the object doesn't move significantly between frames and uses estimation from the previous frame as a guide to the object region on the current one. However, during fast movements, the tracker can lose the target, which requires the detector to re-localize it in the image. MediaPipe Holistic uses pose prediction (on every frame) as an additional ROI prior to reduce the response time of the pipeline when reacting to fast movements. This also enables the model to retain semantic consistency across the body and its parts by preventing a mixup between left and right hands or body parts of one person in the frame with another.

In addition, the resolution of the input frame to the pose model is low enough that the resulting ROIs for face and hands are still too inaccurate to guide the re-cropping of those regions, which require a precise input crop to remain lightweight. To close this accuracy gap we use lightweight face and hand re-crop models that play the role of spatial transformers and cost only ~10% of corresponding model's inference time.

The pipeline is implemented as a MediaPipe graph that uses a holistic landmark subgraph from the holistic landmark module and renders using a dedicated holistic renderer subgraph. The holistic landmark subgraph internally uses a pose landmark module , hand landmark module and face landmark module. Please check them for implementation details.

Note: To visualize a graph, copy the graph and paste it into MediaPipe Visualizer. For more information on how to visualize its associated subgraphs, please see visualizer documentation.

Models

Landmark Models

MediaPipe Holistic utilizes the pose, face and hand landmark models in MediaPipe Pose, MediaPipe Face Mesh and MediaPipe Hands respectively to generate a total of 543 landmarks (33 pose landmarks, 468 face landmarks, and 21 hand landmarks per hand).

Hand Recrop Model

For cases when the accuracy of the pose model is low enough that the resulting ROIs for hands are still too inaccurate we run the additional lightweight hand re-crop model that play the role of spatial transformer and cost only ~10% of hand model inference time.

Solution APIs

Cross-platform Configuration Options

Naming style and availability may differ slightly across platforms/languages.

static_image_mode

If set to false, the solution treats the input images as a video stream. It will try to detect the most prominent person in the very first images, and upon a successful detection further localizes the pose and other landmarks. In subsequent images, it then simply tracks those landmarks without invoking another detection until it loses track, on reducing computation and latency. If set to true, person detection runs every input image, ideal for processing a batch of static, possibly unrelated, images. Default to false.

model_complexity

Complexity of the pose landmark model: 0, 1 or 2. Landmark accuracy as well as inference latency generally go up with the model complexity. Default to 1.

smooth_landmarks

If set to true, the solution filters pose landmarks across different input images to reduce jitter, but ignored if static_image_mode is also set to true. Default to true.

min_detection_confidence

Minimum confidence value ([0.0, 1.0]) from the person-detection model for the detection to be considered successful. Default to 0.5.

min_tracking_confidence

Minimum confidence value ([0.0, 1.0]) from the landmark-tracking model for the pose landmarks to be considered tracked successfully, or otherwise person detection will be invoked automatically on the next input image. Setting it to a higher value can increase robustness of the solution, at the expense of a higher latency. Ignored if static_image_mode is true, where person detection simply runs on every image. Default to 0.5.

Output

Naming style may differ slightly across platforms/languages.

pose_landmarks

A list of pose landmarks. Each landmark consists of the following:

x and y: Landmark coordinates normalized to [0.0, 1.0] by the image width and height respectively.
z: Should be discarded as currently the model is not fully trained to predict depth, but this is something on the roadmap.
visibility: A value in [0.0, 1.0] indicating the likelihood of the landmark being visible (present and not occluded) in the image.

face_landmarks

A list of 468 face landmarks. Each landmark consists of x, y and z. x and y are normalized to [0.0, 1.0] by the image width and height respectively. z represents the landmark depth with the depth at center of the head being the origin, and the smaller the value the closer the landmark is to the camera. The magnitude of z uses roughly the same scale as x.

left_hand_landmarks

A list of 21 hand landmarks on the left hand. Each landmark consists of x, y and z. x and y are normalized to [0.0, 1.0] by the image width and height respectively. z represents the landmark depth with the depth at the wrist being the origin, and the smaller the value the closer the landmark is to the camera. The magnitude of z uses roughly the same scale as x.

right_hand_landmarks

A list of 21 hand landmarks on the right hand, in the same representation as left_hand_landmarks.

Python Solution API

Please first follow general instructions to install MediaPipe Python package, then learn more in the companion Python Colab and the following usage example.

Supported configuration options:

static_image_mode
model_complexity
smooth_landmarks
min_detection_confidence
min_tracking_confidence

import cv2
import mediapipe as mp
mp_drawing = mp.solutions.drawing_utils
mp_holistic = mp.solutions.holistic

# For static images:
with mp_holistic.Holistic(
    static_image_mode=True,
    model_complexity=2) as holistic:
  for idx, file in enumerate(file_list):
    image = cv2.imread(file)
    image_height, image_width, _ = image.shape
    # Convert the BGR image to RGB before processing.
    results = holistic.process(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))

    if results.pose_landmarks:
      print(
          f'Nose coordinates: ('
          f'{results.pose_landmarks.landmark[mp_holistic.PoseLandmark.NOSE].x * image_width}, '
          f'{results.pose_landmarks.landmark[mp_holistic.PoseLandmark.NOSE].y * image_height})'
      )
    # Draw pose, left and right hands, and face landmarks on the image.
    annotated_image = image.copy()
    mp_drawing.draw_landmarks(
        annotated_image, results.face_landmarks, mp_holistic.FACE_CONNECTIONS)
    mp_drawing.draw_landmarks(
        annotated_image, results.left_hand_landmarks, mp_holistic.HAND_CONNECTIONS)
    mp_drawing.draw_landmarks(
        annotated_image, results.right_hand_landmarks, mp_holistic.HAND_CONNECTIONS)
    mp_drawing.draw_landmarks(
        annotated_image, results.pose_landmarks, mp_holistic.POSE_CONNECTIONS)
    cv2.imwrite('/tmp/annotated_image' + str(idx) + '.png', annotated_image)

# For webcam input:
cap = cv2.VideoCapture(0)
with mp_holistic.Holistic(
    min_detection_confidence=0.5,
    min_tracking_confidence=0.5) as holistic:
  while cap.isOpened():
    success, image = cap.read()
    if not success:
      print("Ignoring empty camera frame.")
      # If loading a video, use 'break' instead of 'continue'.
      continue

    # Flip the image horizontally for a later selfie-view display, and convert
    # the BGR image to RGB.
    image = cv2.cvtColor(cv2.flip(image, 1), cv2.COLOR_BGR2RGB)
    # To improve performance, optionally mark the image as not writeable to
    # pass by reference.
    image.flags.writeable = False
    results = holistic.process(image)

    # Draw landmark annotation on the image.
    image.flags.writeable = True
    image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
    mp_drawing.draw_landmarks(
        image, results.face_landmarks, mp_holistic.FACE_CONNECTIONS)
    mp_drawing.draw_landmarks(
        image, results.left_hand_landmarks, mp_holistic.HAND_CONNECTIONS)
    mp_drawing.draw_landmarks(
        image, results.right_hand_landmarks, mp_holistic.HAND_CONNECTIONS)
    mp_drawing.draw_landmarks(
        image, results.pose_landmarks, mp_holistic.POSE_CONNECTIONS)
    cv2.imshow('MediaPipe Holistic', image)
    if cv2.waitKey(5) & 0xFF == 27:
      break
cap.release()

JavaScript Solution API

Please first see general introduction on MediaPipe in JavaScript, then learn more in the companion web demo and the following usage example.