talksik/mediapipe
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Project import generated by Copybara.

Browse Source 
GitOrigin-RevId: ff83882955f1a1e2a043ff4e71278be9d7217bbe
This commit is contained in:
MediaPipe Team
2021-05-05 14:56:16 -04:00
committed by chuoling
parent ecb5b5f44a
commit a9b643e0f5
210 changed files with 5312 additions and 3838 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/framework_concepts/calculators.md
+2 -3
View File
@@ -187,7 +187,7 @@ node {
```
In the calculator implementation, inputs and outputs are also identified by tag
name and index number. In the function below input are output are identified:
name and index number. In the function below input and output are identified:
* By index number: The combined input stream is identified simply by index
`0`.
@@ -355,7 +355,6 @@ class PacketClonerCalculator : public CalculatorBase {
current_[i].At(cc->InputTimestamp()));
// Add a packet to output stream of index i a packet from inputstream i
// with timestamp common to all present inputs
//
} else {
cc->Outputs().Index(i).SetNextTimestampBound(
cc->InputTimestamp().NextAllowedInStream());
@@ -382,7 +381,7 @@ defined your calculator class, register it with a macro invocation
REGISTER_CALCULATOR(calculator_class_name).
Below is a trivial MediaPipe graph that has 3 input streams, 1 node
(PacketClonerCalculator) and 3 output streams.
(PacketClonerCalculator) and 2 output streams.
```proto
input_stream: "room_mic_signal"
docs/framework_concepts/graphs.md
+2 -2
View File
@@ -83,12 +83,12 @@ Below is an example of how to create a subgraph named `TwoPassThroughSubgraph`.
output_stream: "out3"
node {
calculator: "PassThroughculator"
calculator: "PassThroughCalculator"
input_stream: "out1"
output_stream: "out2"
}
node {
calculator: "PassThroughculator"
calculator: "PassThroughCalculator"
input_stream: "out2"
output_stream: "out3"
}
docs/getting_started/android.md
+2 -2
View File
@@ -57,7 +57,7 @@ Please verify all the necessary packages are installed.
* Android SDK Build-Tools 28 or 29
* Android SDK Platform-Tools 28 or 29
* Android SDK Tools 26.1.1
* Android NDK 17c or above
* Android NDK 19c or above
### Option 1: Build with Bazel in Command Line
@@ -111,7 +111,7 @@ app:
* Verify that Android SDK Build-Tools 28 or 29 is installed.
* Verify that Android SDK Platform-Tools 28 or 29 is installed.
* Verify that Android SDK Tools 26.1.1 is installed.
* Verify that Android NDK 17c or above is installed.
* Verify that Android NDK 19c or above is installed.
* Take note of the Android NDK Location, e.g.,
`/usr/local/home/Android/Sdk/ndk-bundle` or
`/usr/local/home/Android/Sdk/ndk/20.0.5594570`.
docs/getting_started/android_archive_library.md
+22 -30
View File
@@ -37,7 +37,7 @@ each project.
load("//mediapipe/java/com/google/mediapipe:mediapipe_aar.bzl", "mediapipe_aar")
mediapipe_aar(
name = "mp_face_detection_aar",
name = "mediapipe_face_detection",
calculators = ["//mediapipe/graphs/face_detection:mobile_calculators"],
)
```
@@ -45,26 +45,29 @@ each project.
2. Run the Bazel build command to generate the AAR.
```bash
bazel build -c opt --host_crosstool_top=@bazel_tools//tools/cpp:toolchain \
--fat_apk_cpu=arm64-v8a,armeabi-v7a --strip=ALWAYS \
//path/to/the/aar/build/file:aar_name
bazel build -c opt --strip=ALWAYS \
--host_crosstool_top=@bazel_tools//tools/cpp:toolchain \
--fat_apk_cpu=arm64-v8a,armeabi-v7a \
//path/to/the/aar/build/file:aar_name.aar
```
For the face detection AAR target we made in the step 1, run:
For the face detection AAR target we made in step 1, run:
```bash
bazel build -c opt --host_crosstool_top=@bazel_tools//tools/cpp:toolchain --fat_apk_cpu=arm64-v8a,armeabi-v7a \
//mediapipe/examples/android/src/java/com/google/mediapipe/apps/aar_example:mp_face_detection_aar
bazel build -c opt --strip=ALWAYS \
--host_crosstool_top=@bazel_tools//tools/cpp:toolchain \
--fat_apk_cpu=arm64-v8a,armeabi-v7a \
//mediapipe/examples/android/src/java/com/google/mediapipe/apps/aar_example:mediapipe_face_detection.aar
# It should print:
# Target //mediapipe/examples/android/src/java/com/google/mediapipe/apps/aar_example:mp_face_detection_aar up-to-date:
# bazel-bin/mediapipe/examples/android/src/java/com/google/mediapipe/apps/aar_example/mp_face_detection_aar.aar
# Target //mediapipe/examples/android/src/java/com/google/mediapipe/apps/aar_example:mediapipe_face_detection.aar up-to-date:
# bazel-bin/mediapipe/examples/android/src/java/com/google/mediapipe/apps/aar_example/mediapipe_face_detection.aar
```
3. (Optional) Save the AAR to your preferred location.
```bash
cp bazel-bin/mediapipe/examples/android/src/java/com/google/mediapipe/apps/aar_example/mp_face_detection_aar.aar
cp bazel-bin/mediapipe/examples/android/src/java/com/google/mediapipe/apps/aar_example/mediapipe_face_detection.aar
/absolute/path/to/your/preferred/location
```
@@ -75,7 +78,7 @@ each project.
2. Copy the AAR into app/libs.
```bash
cp bazel-bin/mediapipe/examples/android/src/java/com/google/mediapipe/apps/aar_example/mp_face_detection_aar.aar
cp bazel-bin/mediapipe/examples/android/src/java/com/google/mediapipe/apps/aar_example/mediapipe_face_detection.aar
/path/to/your/app/libs/
```
@@ -92,29 +95,14 @@ each project.
[the face detection tflite model](https://github.com/google/mediapipe/tree/master/mediapipe/modules/face_detection/face_detection_front.tflite).
```bash
bazel build -c opt mediapipe/mediapipe/graphs/face_detection:mobile_gpu_binary_graph
cp bazel-bin/mediapipe/graphs/face_detection/mobile_gpu.binarypb /path/to/your/app/src/main/assets/
bazel build -c opt mediapipe/graphs/face_detection:face_detection_mobile_gpu_binary_graph
cp bazel-bin/mediapipe/graphs/face_detection/face_detection_mobile_gpu.binarypb /path/to/your/app/src/main/assets/
cp mediapipe/modules/face_detection/face_detection_front.tflite /path/to/your/app/src/main/assets/
```
![Screenshot](../images/mobile/assets_location.png)
4. Make app/src/main/jniLibs and copy OpenCV JNI libraries into
app/src/main/jniLibs.
MediaPipe depends on OpenCV, you will need to copy the precompiled OpenCV so
files into app/src/main/jniLibs. You can download the official OpenCV
Android SDK from
[here](https://github.com/opencv/opencv/releases/download/3.4.3/opencv-3.4.3-android-sdk.zip)
and run:
```bash
cp -R ~/Downloads/OpenCV-android-sdk/sdk/native/libs/arm* /path/to/your/app/src/main/jniLibs/
```
![Screenshot](../images/mobile/android_studio_opencv_location.png)
5. Modify app/build.gradle to add MediaPipe dependencies and MediaPipe AAR.
4. Modify app/build.gradle to add MediaPipe dependencies and MediaPipe AAR.
```
dependencies {
@@ -136,10 +124,14 @@ each project.
implementation "androidx.camera:camera-core:$camerax_version"
implementation "androidx.camera:camera-camera2:$camerax_version"
implementation "androidx.camera:camera-lifecycle:$camerax_version"
// AutoValue
def auto_value_version = "1.6.4"
implementation "com.google.auto.value:auto-value-annotations:$auto_value_version"
annotationProcessor "com.google.auto.value:auto-value:$auto_value_version"
}
```
6. Follow our Android app examples to use MediaPipe in Android Studio for your
5. Follow our Android app examples to use MediaPipe in Android Studio for your
use case. If you are looking for an example, a face detection example can be
found
[here](https://github.com/jiuqiant/mediapipe_face_detection_aar_example) and
docs/getting_started/install.md
+2 -2
View File
@@ -471,7 +471,7 @@ next section.
4. Install Visual C++ Build Tools 2019 and WinSDK
Go to
[the VisualStudio website](ttps://visualstudio.microsoft.com/visual-cpp-build-tools),
[the VisualStudio website](https://visualstudio.microsoft.com/visual-cpp-build-tools),
download build tools, and install Microsoft Visual C++ 2019 Redistributable
and Microsoft Build Tools 2019.
@@ -738,7 +738,7 @@ common build issues.
root@bca08b91ff63:/mediapipe# bash ./setup_android_sdk_and_ndk.sh
# Should print:
# Android NDK is now installed. Consider setting $ANDROID_NDK_HOME environment variable to be /root/Android/Sdk/ndk-bundle/android-ndk-r18b
# Android NDK is now installed. Consider setting $ANDROID_NDK_HOME environment variable to be /root/Android/Sdk/ndk-bundle/android-ndk-r19c
# Set android_ndk_repository and android_sdk_repository in WORKSPACE
# Done
docs/getting_started/python.md
+1 -1
View File
@@ -26,7 +26,7 @@ You can, for instance, activate a Python virtual environment:
$ python3 -m venv mp_env && source mp_env/bin/activate
```
Install MediaPipe Python package and start Python intepreter:
Install MediaPipe Python package and start Python interpreter:
```bash
(mp_env)$ pip install mediapipe
docs/getting_started/troubleshooting.md
+43
View File
@@ -97,6 +97,49 @@ linux_opencv/macos_opencv/windows_opencv.BUILD files for your local opencv
libraries. [This GitHub issue](https://github.com/google/mediapipe/issues/666)
may also help.
## Python pip install failure
The error message:
```
ERROR: Could not find a version that satisfies the requirement mediapipe
ERROR: No matching distribution found for mediapipe
```
after running `pip install mediapipe` usually indicates that there is no qualified MediaPipe Python for your system.
Please note that MediaPipe Python PyPI officially supports the **64-bit**
version of Python 3.7 and above on the following OS:
- x86_64 Linux
- x86_64 macOS 10.15+
- amd64 Windows
If the OS is currently supported and you still see this error, please make sure
that both the Python and pip binary are for Python 3.7 and above. Otherwise,
please consider building the MediaPipe Python package locally by following the
instructions [here](python.md#building-mediapipe-python-package).
## Python DLL load failure on Windows
The error message:
```
ImportError: DLL load failed: The specified module could not be found
```
usually indicates that the local Windows system is missing Visual C++
redistributable packages and/or Visual C++ runtime DLLs. This can be solved by
either installing the official
[vc_redist.x64.exe](https://support.microsoft.com/en-us/topic/the-latest-supported-visual-c-downloads-2647da03-1eea-4433-9aff-95f26a218cc0)
or installing the "msvc-runtime" Python package by running
```bash
$ python -m pip install msvc-runtime
```
Please note that the "msvc-runtime" Python package is not released or maintained
by Microsoft.
## Native method not found
The error message:
docs/images/mobile/aar_location.png
BIN
View File
Binary file not shown.
Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 35 KiB

After

Width:  |  Height:  |  Size: 34 KiB

docs/images/mobile/android_studio_opencv_location.png
BIN
View File
Binary file not shown.
Side by Side

Before

Width:  |  Height:  |  Size: 75 KiB

docs/images/mobile/assets_location.png
BIN
View File
Binary file not shown.
Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 29 KiB

After

Width:  |  Height:  |  Size: 42 KiB

docs/images/mobile/pose_tracking_example.gif
BIN
View File
Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 2.3 MiB

docs/images/mobile/pose_tracking_upper_body_example.gif
BIN
View File
Binary file not shown.
Side by Side

Before

Width:  |  Height:  |  Size: 6.9 MiB

docs/solutions/face_detection.md
+1 -1
View File
@@ -77,7 +77,7 @@ Supported configuration options:
```python
import cv2
import mediapipe as mp
mp_face_detction = mp.solutions.face_detection
mp_face_detection = mp.solutions.face_detection
mp_drawing = mp.solutions.drawing_utils
# For static images:
docs/solutions/holistic.md
+10 -11
View File
@@ -135,12 +135,11 @@ 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`.
#### upper_body_only
#### model_complexity
If set to `true`, the solution outputs only the 25 upper-body pose landmarks
(535 in total) instead of the full set of 33 pose landmarks (543 in total). Note
that upper-body-only prediction may be more accurate for use cases where the
lower-body parts are mostly out of view. Default to `false`.
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
@@ -207,7 +206,7 @@ install MediaPipe Python package, then learn more in the companion
Supported configuration options:
* [static_image_mode](#static_image_mode)
* [upper_body_only](#upper_body_only)
* [model_complexity](#model_complexity)
* [smooth_landmarks](#smooth_landmarks)
* [min_detection_confidence](#min_detection_confidence)
* [min_tracking_confidence](#min_tracking_confidence)
@@ -219,7 +218,9 @@ mp_drawing = mp.solutions.drawing_utils
mp_holistic = mp.solutions.holistic
# For static images:
with mp_holistic.Holistic(static_image_mode=True) as holistic:
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
@@ -240,8 +241,6 @@ with mp_holistic.Holistic(static_image_mode=True) as holistic:
annotated_image, results.left_hand_landmarks, mp_holistic.HAND_CONNECTIONS)
mp_drawing.draw_landmarks(
annotated_image, results.right_hand_landmarks, mp_holistic.HAND_CONNECTIONS)
# Use mp_holistic.UPPER_BODY_POSE_CONNECTIONS for drawing below when
# upper_body_only is set to True.
mp_drawing.draw_landmarks(
annotated_image, results.pose_landmarks, mp_holistic.POSE_CONNECTIONS)
cv2.imwrite('/tmp/annotated_image' + str(idx) + '.png', annotated_image)
@@ -291,7 +290,7 @@ and the following usage example.
Supported configuration options:
* [upperBodyOnly](#upper_body_only)
* [modelComplexity](#model_complexity)
* [smoothLandmarks](#smooth_landmarks)
* [minDetectionConfidence](#min_detection_confidence)
* [minTrackingConfidence](#min_tracking_confidence)
@@ -348,7 +347,7 @@ const holistic = new Holistic({locateFile: (file) => {
return `https://cdn.jsdelivr.net/npm/@mediapipe/holistic/${file}`;
}});
holistic.setOptions({
upperBodyOnly: false,
modelComplexity: 1,
smoothLandmarks: true,
minDetectionConfidence: 0.5,
minTrackingConfidence: 0.5
docs/solutions/models.md
+6 -6
View File
@@ -15,10 +15,10 @@ nav_order: 30
### [Face Detection](https://google.github.io/mediapipe/solutions/face_detection)
* Face detection model for front-facing/selfie camera:
[TFLite model](https://github.com/google/mediapipe/tree/master/mediapipe/models/face_detection_front.tflite),
[TFLite model](https://github.com/google/mediapipe/tree/master/mediapipe/modules/face_detection/face_detection_front.tflite),
[TFLite model quantized for EdgeTPU/Coral](https://github.com/google/mediapipe/tree/master/mediapipe/examples/coral/models/face-detector-quantized_edgetpu.tflite)
* Face detection model for back-facing camera:
[TFLite model ](https://github.com/google/mediapipe/tree/master/mediapipe/models/face_detection_back.tflite)
[TFLite model ](https://github.com/google/mediapipe/tree/master/mediapipe/modules/face_detection/face_detection_back.tflite)
* [Model card](https://mediapipe.page.link/blazeface-mc)
### [Face Mesh](https://google.github.io/mediapipe/solutions/face_mesh)
@@ -49,10 +49,10 @@ nav_order: 30
* Pose detection model:
[TFLite model](https://github.com/google/mediapipe/tree/master/mediapipe/modules/pose_detection/pose_detection.tflite)
* Full-body pose landmark model:
[TFLite model](https://github.com/google/mediapipe/tree/master/mediapipe/modules/pose_landmark/pose_landmark_full_body.tflite)
* Upper-body pose landmark model:
[TFLite model](https://github.com/google/mediapipe/tree/master/mediapipe/modules/pose_landmark/pose_landmark_upper_body.tflite)
* Pose landmark model:
[TFLite model (lite)](https://github.com/google/mediapipe/tree/master/mediapipe/modules/pose_landmark/pose_landmark_lite.tflite),
[TFLite model (full)](https://github.com/google/mediapipe/tree/master/mediapipe/modules/pose_landmark/pose_landmark_full.tflite),
[TFLite model (heavy)](https://github.com/google/mediapipe/tree/master/mediapipe/modules/pose_landmark/pose_landmark_heavy.tflite)
* [Model card](https://mediapipe.page.link/blazepose-mc)
### [Holistic](https://google.github.io/mediapipe/solutions/holistic)
docs/solutions/pose.md
+33 -47
View File
@@ -30,8 +30,7 @@ overlay of digital content and information on top of the physical world in
augmented reality.
MediaPipe Pose is a ML solution for high-fidelity body pose tracking, inferring
33 3D landmarks on the whole body (or 25 upper-body landmarks) from RGB video
frames utilizing our
33 3D landmarks on the whole body from RGB video frames utilizing our
[BlazePose](https://ai.googleblog.com/2020/08/on-device-real-time-body-pose-tracking.html)
research that also powers the
[ML Kit Pose Detection API](https://developers.google.com/ml-kit/vision/pose-detection).
@@ -40,9 +39,9 @@ environments for inference, whereas our method achieves real-time performance on
most modern [mobile phones](#mobile), [desktops/laptops](#desktop), in
[python](#python-solution-api) and even on the [web](#javascript-solution-api).
![pose_tracking_upper_body_example.gif](../images/mobile/pose_tracking_upper_body_example.gif) |
:--------------------------------------------------------------------------------------------: |
*Fig 1. Example of MediaPipe Pose for upper-body pose tracking.* |
![pose_tracking_example.gif](../images/mobile/pose_tracking_example.gif) |
:----------------------------------------------------------------------: |
*Fig 1. Example of MediaPipe Pose for pose tracking.* |
## ML Pipeline
@@ -77,6 +76,23 @@ Note: To visualize a graph, copy the graph and paste it into
to visualize its associated subgraphs, please see
[visualizer documentation](../tools/visualizer.md).
## Pose Estimation Quality
To evaluate the quality of our [models](./models.md#pose) against other
well-performing publicly available solutions, we use a validation dataset,
consisting of 1k images with diverse Yoga, HIIT, and Dance postures. Each image
contains only a single person located 2-4 meters from the camera. To be
consistent with other solutions, we perform evaluation only for 17 keypoints
from [COCO topology](https://cocodataset.org/#keypoints-2020).
Method | [mAP](https://cocodataset.org/#keypoints-eval) | [[email protected]](https://github.com/cbsudux/Human-Pose-Estimation-101) | [FPS](https://en.wikipedia.org/wiki/Frame_rate), Pixel 3 [TFLite GPU](https://www.tensorflow.org/lite/performance/gpu_advanced) | [FPS](https://en.wikipedia.org/wiki/Frame_rate), MacBook Pro (15-inch, 2017)
----------------------------------------------------------------------------------------------------- | ---------------------------------------------: | --------------------------------------------------------------: | ------------------------------------------------------------------------------------------------------------------------------: | ---------------------------------------------------------------------------:
BlazePose.Lite | 49.1 | 91.7 | 49 | 40
BlazePose.Full | 64.5 | 95.8 | 40 | 37
BlazePose.Heavy | 70.9 | 97.0 | 19 | 26
[AlphaPose.ResNet50](https://github.com/MVIG-SJTU/AlphaPose) | 57.6 | 93.1 | N/A | N/A
[Apple Vision](https://developer.apple.com/documentation/vision/detecting_human_body_poses_in_images) | 37.0 | 85.3 | N/A | N/A
## Models
### Person/pose Detection Model (BlazePose Detector)
@@ -97,11 +113,8 @@ hip midpoints.
### Pose Landmark Model (BlazePose GHUM 3D)
The landmark model in MediaPipe Pose comes in two versions: a full-body model
that predicts the location of 33 pose landmarks (see figure below), and an
upper-body version that only predicts the first 25. The latter may be more
accurate than the former in scenarios where the lower-body parts are mostly out
of view.
The landmark model in MediaPipe Pose predicts the location of 33 pose landmarks
(see figure below).
Please find more detail in the
[BlazePose Google AI Blog](https://ai.googleblog.com/2020/08/on-device-real-time-body-pose-tracking.html),
@@ -129,12 +142,11 @@ 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`.
#### upper_body_only
#### model_complexity
If set to `true`, the solution outputs only the 25 upper-body pose landmarks.
Otherwise, it outputs the full set of 33 pose landmarks. Note that
upper-body-only prediction may be more accurate for use cases where the
lower-body parts are mostly out of view. Default to `false`.
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
@@ -170,9 +182,6 @@ A list of pose landmarks. Each lanmark consists of the following:
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`.
Note: `z` is predicted only in full-body mode, and should be discarded when
[upper_body_only](#upper_body_only) is `true`.
* `visibility`: A value in `[0.0, 1.0]` indicating the likelihood of the
landmark being visible (present and not occluded) in the image.
@@ -185,7 +194,7 @@ install MediaPipe Python package, then learn more in the companion
Supported configuration options:
* [static_image_mode](#static_image_mode)
* [upper_body_only](#upper_body_only)
* [model_complexity](#model_complexity)
* [smooth_landmarks](#smooth_landmarks)
* [min_detection_confidence](#min_detection_confidence)
* [min_tracking_confidence](#min_tracking_confidence)
@@ -198,7 +207,9 @@ mp_pose = mp.solutions.pose
# For static images:
with mp_pose.Pose(
static_image_mode=True, min_detection_confidence=0.5) as pose:
static_image_mode=True,
model_complexity=2,
min_detection_confidence=0.5) as pose:
for idx, file in enumerate(file_list):
image = cv2.imread(file)
image_height, image_width, _ = image.shape
@@ -214,8 +225,6 @@ with mp_pose.Pose(
)
# Draw pose landmarks on the image.
annotated_image = image.copy()
# Use mp_pose.UPPER_BODY_POSE_CONNECTIONS for drawing below when
# upper_body_only is set to True.
mp_drawing.draw_landmarks(
annotated_image, results.pose_landmarks, mp_pose.POSE_CONNECTIONS)
cv2.imwrite('/tmp/annotated_image' + str(idx) + '.png', annotated_image)
@@ -259,7 +268,7 @@ and the following usage example.
Supported configuration options:
* [upperBodyOnly](#upper_body_only)
* [modelComplexity](#model_complexity)
* [smoothLandmarks](#smooth_landmarks)
* [minDetectionConfidence](#min_detection_confidence)
* [minTrackingConfidence](#min_tracking_confidence)
@@ -306,7 +315,7 @@ const pose = new Pose({locateFile: (file) => {
return `https://cdn.jsdelivr.net/npm/@mediapipe/pose/${file}`;
}});
pose.setOptions({
upperBodyOnly: false,
modelComplexity: 1,
smoothLandmarks: true,
minDetectionConfidence: 0.5,
minTrackingConfidence: 0.5
@@ -347,16 +356,6 @@ to visualize its associated subgraphs, please see
* iOS target:
[`mediapipe/examples/ios/posetrackinggpu:PoseTrackingGpuApp`](http:/mediapipe/examples/ios/posetrackinggpu/BUILD)
#### Upper-body Only
* Graph:
[`mediapipe/graphs/pose_tracking/upper_body_pose_tracking_gpu.pbtxt`](https://github.com/google/mediapipe/tree/master/mediapipe/graphs/pose_tracking/upper_body_pose_tracking_gpu.pbtxt)
* Android target:
[(or download prebuilt ARM64 APK)](https://drive.google.com/file/d/1uKc6T7KSuA0Mlq2URi5YookHu0U3yoh_/view?usp=sharing)
[`mediapipe/examples/android/src/java/com/google/mediapipe/apps/upperbodyposetrackinggpu:upperbodyposetrackinggpu`](https://github.com/google/mediapipe/tree/master/mediapipe/examples/android/src/java/com/google/mediapipe/apps/upperbodyposetrackinggpu/BUILD)
* iOS target:
[`mediapipe/examples/ios/upperbodyposetrackinggpu:UpperBodyPoseTrackingGpuApp`](http:/mediapipe/examples/ios/upperbodyposetrackinggpu/BUILD)
### Desktop
Please first see general instructions for [desktop](../getting_started/cpp.md)
@@ -375,19 +374,6 @@ on how to build MediaPipe examples.
* Target:
[`mediapipe/examples/desktop/pose_tracking:pose_tracking_gpu`](https://github.com/google/mediapipe/tree/master/mediapipe/examples/desktop/pose_tracking/BUILD)
#### Upper-body Only
* Running on CPU
* Graph:
[`mediapipe/graphs/pose_tracking/upper_body_pose_tracking_cpu.pbtxt`](https://github.com/google/mediapipe/tree/master/mediapipe/graphs/pose_tracking/upper_body_pose_tracking_cpu.pbtxt)
* Target:
[`mediapipe/examples/desktop/upper_body_pose_tracking:upper_body_pose_tracking_cpu`](https://github.com/google/mediapipe/tree/master/mediapipe/examples/desktop/upper_body_pose_tracking/BUILD)
* Running on GPU
* Graph:
[`mediapipe/graphs/pose_tracking/upper_body_pose_tracking_gpu.pbtxt`](https://github.com/google/mediapipe/tree/master/mediapipe/graphs/pose_tracking/upper_body_pose_tracking_gpu.pbtxt)
* Target:
[`mediapipe/examples/desktop/upper_body_pose_tracking:upper_body_pose_tracking_gpu`](https://github.com/google/mediapipe/tree/master/mediapipe/examples/desktop/upper_body_pose_tracking/BUILD)
## Resources
* Google AI Blog: