Environment

Running Mediapipe on GPU with OpenGL ES < 3.0

When I run below command bazel build -c opt --define MEDIAPIPE_DISABLE_GPU=1 --action_env PYTHON_BIN_PATH="C://Python//Python37//python.exe" mediapipe/examples/desktop/hello_world it generates the error:
ERROR: D:/git/mediapipe/mediapipe/framework/BUILD:824:1: C++ compilation of rule '//mediapipe/framework:packet' failed (Exit 2)
.\mediapipe/framework/packet.h(448): error C2672: 'absl::lts_2020_02_25::make_unique': no matching overloaded function found
.\mediapipe/framework/packet.h(448): error C2783: 'void std::make_unique(_Types &&...)': could not deduce template argument for '__formal'
C:\Program Files (x86)\Microsoft Visual Studio\2019\BuildTools\VC\Tools\MSVC\14.26.28801\include\memory(2078): note: see declaration of 'std::make_unique'
Target //mediapipe/examples/desktop/hello_world:hello_world failed to build
Use --verbose_failures to see the command lines of failed build steps.
INFO: Elapsed time: 71.407s, Critical Path: 68.40s
INFO: 152 processes: 152 local.
FAILED: Build did NOT complete successfully
I am using C++ BuildTools version 14.26.28801 If I rollback to May version, it builds successfully.
How do you rollback to May version? Do you have a link?
https://github.com/google/mediapipe/releases/tag/v0.7.5 You can download the older version as zip here

Or just do mediapipe> git checkout v0.7.5.

cwlee200

comment created time in 17 days

IssuesEvent

issue closedgoogle/mediapipe

Hi,

I realised after multiple days of testing and deep googling that currently no Virtual Machine supports OpenGL ES 3+. I tested on VirtualBox and VMware Player, no one supports it. As said in your docs, "OpenGL ES 3.1 or greater is required (on Android/Linux systems) for running machine learning inference calculators and graphs." which is what interests me.

Are you aware of any way to bypass the problem ? Or is it in you plans to extend the system requirements ability to OpenGL ES < 3 ?

It surprised me that so little is said on google about it. I don't believe I am the only one to run on virtual machines ...

Thanks for your response !

closed time in 17 days

QBoulanger

issue commentgoogle/mediapipe

Hand detection example fails to build on RPI 4B

The issue is the OpenCV library is not configured correctly for MediaPipe. How do you install the OpenCV library and config it for MediaPipe? We provide 3 options to config OpenCV for MediaPipe in the step 3 of https://google.github.io/mediapipe/getting_started/install#installing-on-debian-and-ubuntu. Moreover, https://github.com/google/mediapipe/issues/666 may also help.

codeNick7

comment created time in 17 days

issue commentgoogle/mediapipe

Build fails handtracking example eigen_archive issue

in mediapipe/framework/packet.h I removed T{}.GetTypeName(), [] { return absl::make_unique<Holder<T>>(new T); })); replaced with T{}.GetTypeName(), absl::make_unique<Holder<T>>));

Thanks for your reply. I did the same thing a few days ago. But, it caused another compilation error when I built the hand tracking demo. I guess it doesn't fix the actual issue.

You can use v0.7.5 (https://github.com/google/mediapipe/tree/v0.7.5) instead. That's the initial release of the Windows support.

rohabimbesat

comment created time in 17 days

issue commentgoogle/mediapipe

Build fails handtracking example eigen_archive issue

Could share how you modified packet.h?

rohabimbesat

comment created time in 19 days

issue commentgoogle/mediapipe

Whats mean Z coordinate ?

The issue occurs at https://github.com/google/mediapipe/blob/e9fbe868e55fa23aaabc31f9f847c22287062850/mediapipe/framework/packet.h#L448

GCC and clang are ok with the syntax [] { return absl::make_unique<Holder<T>>(new T); }) on Linux and macos. However, MSVC is not. It's unclear to me what't the right fix for MSVC on Windows. Any suggestions are welcome.

Please note that MediaPipe Windows support is experimental since we have limited resources and mainly focus on Linux and mobile platforms.

cwlee200

comment created time in 20 days

issue commentgoogle/mediapipe

Does this answer https://github.com/google/mediapipe/issues/742#issuecomment-639104199 help?

www516717402

comment created time in 20 days

issue commentgoogle/mediapipe

Disclaimer: MediaPipe Windows support is experimental since we have limited resources and mainly focus on Linux and mobile platforms.

The issue occurs at https://github.com/google/mediapipe/blob/e9fbe868e55fa23aaabc31f9f847c22287062850/mediapipe/framework/packet.h#L448.

GCC and clang are ok with the syntax [] { return absl::make_unique<Holder<T>>(new T); }) on Linux and macos. However, MSVC is not. It's unclear to me what't the right fix for MSVC on Windows. Any suggestions are welcome.

cwlee200

comment created time in 20 days

issue commentgoogle/mediapipe

Build fails handtracking example eigen_archive issue

The issue occurs at https://github.com/google/mediapipe/blob/e9fbe868e55fa23aaabc31f9f847c22287062850/mediapipe/framework/packet.h#L448.

GCC and clang are ok with the syntax [] { return absl::make_unique<Holder<T>>(new T); }) on Linux and macos. However, MSVC is not. It's unclear what't the right fix for MSVC.

cwlee200

comment created time in 20 days

issue commentgoogle/mediapipe

The error seems to be Stdout: Stderr: /usr/bin/bash: patch: command not found ERROR, which means bazel fails to apply some patches on the eigen_archive library.

Did you modify the codebase or just use the latest version of MediaPipe as is? I am asking because a group of users report that they can't build MediaPipe on Windows with the latest version (https://github.com/google/mediapipe/issues/908).

rohabimbesat

comment created time in 20 days

issue commentgoogle/mediapipe

How can we used mediapipe object tracking with own object detector?

I think MediaPipe's tensorflow inference calculator only supports tensorflow models. If you need to do model inference with some other inference engine, unfortunately MediaPipe doesn't have such calculators. But, you definitely can implement a new inference calculator by yourself in C++. Here is how we implement tensorflow inference calculator: https://github.com/google/mediapipe/blob/61bc4556af586bb3837f17187568d1345b3ee8f5/mediapipe/calculators/tensorflow/tensorflow_inference_calculator.cc

sainisanjay

comment created time in 20 days

issue commentgoogle/mediapipe

Can't Run Hello World Example

Hi, can you show the full log and the build command you use? Seems like @com_google_absl is not properly downloaded, which results to the error '@com_google_absl' can not be resolved.

udogukankurt

comment created time in 22 days

issue commentgoogle/mediapipe

Errors when I build a MediaPipe AAR 在构建MediaPipe AAR包时遇到的报错

Try the following fixes. Note that those fixes may only work on native Windows with MSVC. Still looking for a more general solution that works across platforms (Linux, macOS, Windows, Android, and iOS).

diff --git a/mediapipe/framework/legacy_calculator_support.h b/mediapipe/framework/legacy_calculator_support.h          
index 4cd15ce..34139da 100644                                                                                           
--- a/mediapipe/framework/legacy_calculator_support.h                                                                   
+++ b/mediapipe/framework/legacy_calculator_support.h                                                                   
@@ -73,12 +73,12 @@ class LegacyCalculatorSupport {                                                                      
// definitions must be in the .cc file. However, a declaration still needs to be                                        
// included in the header, or some compilers will assume they have no                                                   
// definition.                                                                                                         
-template <>                                                                                                            
-thread_local CalculatorContext*                                                                                        
-    LegacyCalculatorSupport::Scoped<CalculatorContext>::current_;                                                      -template <>                                                                                                            
-thread_local CalculatorContract*                                                                                       
-    LegacyCalculatorSupport::Scoped<CalculatorContract>::current_;                                                     +// template <>                                                                                                         
+// thread_local CalculatorContext*                                                                                     
+//     LegacyCalculatorSupport::Scoped<CalculatorContext>::current_;                                                   +// template <>                                                                                                         
+// thread_local CalculatorContract*                                                                                    
+//     LegacyCalculatorSupport::Scoped<CalculatorContract>::current_;                                                   
}  // namespace mediapipe                                                                                                                                                                                                                      
diff --git a/mediapipe/framework/packet.h b/mediapipe/framework/packet.h                                                index f6b1774..3f296f9 100644                                                                                           
--- a/mediapipe/framework/packet.h                                                                                      
+++ b/mediapipe/framework/packet.h                                                                                      
@@ -445,7 +445,7 @@ struct MessageRegistrationImpl {                                                                     template <typename T>                                                                                                   NoDestructor<mediapipe::RegistrationToken>                                                                                  MessageRegistrationImpl<T>::registration(MessageHolderRegistry::Register(                                       
-        T{}.GetTypeName(), [] { return absl::make_unique<Holder<T>>(new T); }));                                    
+        T{}.GetTypeName(), absl::make_unique<Holder<T>>));                                                                                                                                                                                      
// For non-Message payloads, this does nothing.                                                                         
template <typename T, typename Enable = void>

cwlee200

comment created time in 22 days

issue commentgoogle/mediapipe

The actual error is "GET returned 403 Forbidden", which apparently is a network issue. Can you verify your network setting?

You probably can find some useful solutions at https://github.com/google/mediapipe/issues/3, https://github.com/google/mediapipe/issues/581, and https://github.com/google/mediapipe/issues/601.

ZZZhaowendao

comment created time in a month

issue closedgoogle/mediapipe

Segmentation fault (core dumped) when running hello world example for desktop on WSL

I've followed the installation steps on WSL. But when I ran the hello world example for desktop with the command I found in MediaPipe docs : bazel run --define MEDIAPIPE_DISABLE_GPU=1
mediapipe/examples/desktop/hello_world:hello_world. The build completed successfully and there is a segmentation fault (core dumped) without any error :

error

Configuration : WSL : Ubuntu 20.04 LTS GPU : Intel UHD Graphics 600 Bazel : 2.0.0 gcc : 9.3.0 g++ : 9.3.0 opencv : 4 ffmpeg : 4.2.2

closed time in a month

laurieban

issue commentgoogle/mediapipe

Segmentation fault (core dumped) when running hello world example for desktop on WSL

The fix is released in https://github.com/google/mediapipe/commit/e9fbe868e55fa23aaabc31f9f847c22287062850.

laurieban

comment created time in a month

issue closedgoogle/mediapipe

Error starting example hello world file (Initial Setup)

I've installed MediaPipe on Ubuntu 20.04 based on the instructions given by the MediaPipe team, though when I try to run the example I keep receiving the following error:

Error

ERROR: Analysis of target '//mediapipe/examples/desktop/hello_world:hello_world' failed; build aborted: Traceback (most recent call last):
 File "/root/.cache/bazel/_bazel_root/6be9f6de2f9f4c6f3d41746e8583a46a/external/org_tensorflow/third_party/py/python_configure.bzl", line 208
  get_python_bin(repository_ctx)
 File "/root/.cache/bazel/_bazel_root/6be9f6de2f9f4c6f3d41746e8583a46a/external/org_tensorflow/third_party/remote_config/common.bzl", line 44, in get_python_bin
  which(repository_ctx, "python")
 File "/root/.cache/bazel/_bazel_root/6be9f6de2f9f4c6f3d41746e8583a46a/external/org_tensorflow/third_party/remote_config/common.bzl", line 29, in which
  execute(repository_ctx, <1 more arguments>)
 File "/root/.cache/bazel/_bazel_root/6be9f6de2f9f4c6f3d41746e8583a46a/external/org_tensorflow/third_party/remote_config/common.bzl", line 208, in execute
  fail(<1 more arguments>)
Repository command failed

I have tried reinstalling python and the entire ubuntu system but still no luck...

I'm running Ubuntu 20.04 on a VM using VirtualBox if that helps(?)

closed time in a month

Mystickal

issue commentgoogle/mediapipe

Error starting example hello world file (Initial Setup)

The fix is released in https://github.com/google/mediapipe/commit/e9fbe868e55fa23aaabc31f9f847c22287062850.

Mystickal

comment created time in a month

push eventgoogle/mediapipe

MediaPipe Team

commit sha e9fbe868e55fa23aaabc31f9f847c22287062850

Project import generated by Copybara. GitOrigin-RevId: e3a43e4e5e519cd14df7095749059e2613bdcf76

view details

push time in a month

push eventgoogle/mediapipe

MediaPipe Team

commit sha e9fbe868e55fa23aaabc31f9f847c22287062850

Project import generated by Copybara. GitOrigin-RevId: e3a43e4e5e519cd14df7095749059e2613bdcf76

view details

push time in a month

issue commentgoogle/mediapipe

ffmpeg_linux.BUILD include path config incorrect

The current ffmpeg still works on my end. Maybe we use different libav versions? My ffmpeg version info is 4.2.2 and libav headers are organized as the following.

$ ffmpeg -version
ffmpeg version 4.2.2-1+build3 Copyright (c) 2000-2019 the FFmpeg developers
built with gcc 9 (Debian 9.2.1-19)

$ ls /usr/include/x86_64-linux-gnu/ | grep libav
libavcodec
libavformat
libavresample
libavutil
$ ls /usr/include/x86_64-linux-gnu/libavcodec/
ac3_parser.h   avcodec.h  avfft.h    dirac.h       dxva2.h  mediacodec.h  vaapi.h  version.h       vorbis_parser.h
adts_parser.h  avdct.h    d3d11va.h  dv_profile.h  jni.h    qsv.h         vdpau.h  videotoolbox.h  xvmc.h

bazaha

comment created time in a month

issue closedgoogle/mediapipe

Where download app(android apk) demo?

Where to download APP(android apk) demo?

closed time in a month

www516717402

issue closedgoogle/mediapipe

How to use pb file conver to tflite file？

I have a saved_model.pb by training in Tensorflow Object Detection API. I can use this model in computer. But now i want to use it in my phone, so i want to make a tflite model. Can you give me some advises, please?

closed time in a month

teafff

issue commentgoogle/mediapipe

How to use pb file conver to tflite file？

TensorFlow can do this for you. Please take a look at https://www.tensorflow.org/lite/convert.

teafff

comment created time in a month

issue closedgoogle/mediapipe

unable to run multi-hand tracking on video

run

GLOG_logtostderr=1 bazel-bin/mediapipe/examples/desktop/hand_tracking/hand_tracking_cpu \
  --calculator_graph_config_file=mediapipe/graphs/hand_tracking/multi_hand_tracking_desktop_live.pbtxt \
  --input_video_path=/test/no_mask_no_teacher.mp4 \
  --output_video_path=/test/1.mp4 \

erro message

[libprotobuf ERROR external/com_google_protobuf/src/google/protobuf/text_format.cc:309] Error parsing text-format mediapipe.CalculatorGraphConfig: 17:75: Could not find type "type.googleapis.com/mediapipe.CollectionHasMinSizeCalculatorOptions" stored in google.protobuf.Any.
F20200707 02:44:45.360370   118 parse_text_proto.h:27] Check failed: google::protobuf::TextFormat::ParseFromString(input, &result) 
*** Check failure stack trace: ***
    @     0x5645eb5304cf  google::LogMessage::SendToLog()
    @     0x5645eb52f19b  google::LogMessage::Flush()
    @     0x5645eb530db9  google::LogMessageFatal::~LogMessageFatal()
    @     0x5645eb152bc5  RunMPPGraph()
    @     0x5645eb145019  main
    @     0x7efd8d0b7b97  __libc_start_main
    @     0x5645eb15104a  _start
    @              (nil)  (unknown)
Aborted (core dumped)

closed time in a month

Chandler-Shuyu-Xie

issue commentgoogle/mediapipe

unable to run multi-hand tracking on video

Because you use the wrong binary.

In docs/solutions/hands.md, we mention that

With Multi-hand Support
Running on CPU
Graph: mediapipe/graphs/hand_tracking/multi_hand_tracking_desktop_live.pbtxt
Target: mediapipe/examples/desktop/multi_hand_tracking:multi_hand_tracking_cpu

Please do

bazel-bin/mediapipe/examples/desktop/multi_hand_tracking/multi_hand_tracking_cpu \
  --calculator_graph_config_file=mediapipe/graphs/hand_tracking/multi_hand_tracking_desktop_live.pbtxt

Chandler-Shuyu-Xie

comment created time in a month

issue closedgoogle/mediapipe

face mesh keypoint

Screenshot from 2020-05-19 14-25-55 i use mediapipe/modules/face_detection/face_detection_front.tflite and mediapipe/modules/face_landmark/face_landmark.tflite to calculate face landmark but keypoint is not correct, possible reason ?

closed time in a month

wangqiang624731186

issue commentgoogle/mediapipe

face mesh keypoint

We are closing this issue for now due to lack of activity.

wangqiang624731186

comment created time in a month

issue closedgoogle/mediapipe

Not a valid win32 application

I am following this tutorial: https://google.github.io/mediapipe/getting_started/building_examples.html for Android. However, when I try either method (with or without Android Studio) I am getting an error with bazel not being recognized as an application:

Error:Cannot run program "\wsl$\Ubuntu\usr\bin\bazel-2.0.0" (in directory "\wsl$\Ubuntu\mediapipe"): CreateProcess error=193, %1 is not a valid Win32 application

I followed this tutorial to install Bazel: https://docs.bazel.build/versions/master/install-ubuntu.html because the recommended installation method below did not work:

username@DESKTOP-TMVLBJ:
~$ curl -sLO --retry 5 --retry-max-time 10
https://storage.googleapis.com/bazel/2.0.0/release/bazel-2.0.0-installer-linux-x86_64.sh &&
sudo mkdir -p /usr/local/bazel/2.0.0 &&
chmod 755 bazel-2.0.0-installer-linux-x86_64.sh &&
sudo ./bazel-2.0.0-installer-linux-x86_64.sh --prefix=/usr/local/bazel/2.0.0 &&
source /usr/local/bazel/2.0.0/lib/bazel/bin/bazel-complete.bash username@DESKTOP-TMVLBJ1:~$ /usr/local/bazel/2.0.0/lib/bazel/bin/bazel version &&
alias bazel='/usr/local/bazel/2.0.0/lib/bazel/bin/bazel'

I am running this on Ubuntu 20 but the Bazel docs say it mainly supports v16 and v18, but should likely work on other releases. This is the only issue I can think of, so I would appreciate any suggestions that could help me fix this issue.

closed time in a month

j-vent

issue commentgoogle/mediapipe

Not a valid win32 application

We are closing this issue for now due to lack of activity.

j-vent

comment created time in a month

issue closedgoogle/mediapipe

youtube8m feature extractor too slow to use

I try to use youtube8m feature extractor to extract audio feature of video. But it is too slow (about 10s/video) because it is only use CPU. Is there any way to speed up (such as using GPU)?

closed time in a month

john159151

issue closedgoogle/mediapipe

Object tracking box info

Hello, I followed the instructions regarding setting up Objecttracking gpu on the android device. It is successfully running. However, I want get the tracking box info (Rect, label, Id , confidence). Could you please help me out here? I tried to add the following code on OnCreate the app starts crashing. processor.addPacketCallback("tracked_detections",(packet)->{ byte[] box = PacketGetter.getProtoBytes(packet); // then check if it is null or not }); Could you please give me some insights to solve this issue? Thank you in advance

closed time in a month

WildSpirit007

issue commentgoogle/mediapipe

Object tracking box info

We are closing this issue for now due to lack of activity.

WildSpirit007

comment created time in a month

issue commentgoogle/mediapipe

C++ compilation error when building mp_multi_hand_aar

Unfortunately, the log doesn't show the error details. If you have time, could you append --verbose_failures to the build command to let it print the full command line for commands that failed and report back?

AlexYiningLiu

comment created time in a month

issue commentgoogle/mediapipe

[Question] Where is the saved model of KNIFT?

@wzc11 will answer your question soon.

dinfer

comment created time in a month

issue closedgoogle/mediapipe

Hi,

I started a project using MediaPipe a few weeks ago and everything was fine until I had to clean install my Ubuntu (18.04) for a separate issue.

Bazel: 3.3.0 Ubuntu: 18.0.4 Android studio: 4.0 SDK: 30 NDK: 19.2.5345600

Steps:

The "Hello World" example (bazel run --copt -DMESA_EGL_NO_X11_HEADERS --copt -DEGL_NO_X11
mediapipe/examples/desktop/hello_world:hello_world) seems to work okay.
Bazel -> Sync -> Sync project with BUILD files also finishes with no errors
Run -> Using the edit configurations in the guide produces the error:

[Prepa] Installing //mediapipe/examples/android/src/java/com/google/mediapipe/apps/robotcontrol:mainapp
ERROR: <PATH>/mediapipe/mediapipe/examples/android/src/java/com/google/mediapipe/apps/robotcontrol/BUILD:62:15: Installing //mediapipe/examples/android/src/java/com/google/mediapipe/apps/robotcontrol:mainapp failed (Exit 1)
Error:
adb command: ['/home/mikroszolos/Android/Sdk/platform-tools/adb', '-s', 'c5e5add5', '-s', 'c5e5add5:tcp:5037', '-s', 'c5e5add5', '-s', 'c5e5add5:tcp:5037', 'shell', 'mkdir -p /data/local/tmp/incrementaldeployment/com.google.mediapipe.apps.robotcontrol/dex']
return code: 1
stdout: 
stderr: adb: unknown host service
----------------
Note: The failure of target @bazel_tools//tools/android:incremental_install (with exit code 1) may have been caused by the fact that it is a Python 2 program that was built in the host configuration, which uses Python 3. You can change the host configuration (for the entire build) to instead use Python 2 by setting --host_force_python=PY2.

If this error started occurring in Bazel 0.27 and later, it may be because the Python toolchain now enforces that targets analyzed as PY2 and PY3 run under a Python 2 and Python 3 interpreter, respectively. See https://github.com/bazelbuild/bazel/issues/7899 for more information.
----------------
Target //mediapipe/examples/android/src/java/com/google/mediapipe/apps/robotcontrol:mainapp failed to build
Use --verbose_failures to see the command lines of failed build steps.
INFO: Elapsed time: 5.366s, Critical Path: 0.38s
INFO: 0 processes.
FAILED: Build did NOT complete successfully
FAILED: Build did NOT complete successfully
INFO: Build Event Protocol files produced successfully.
FAILED: Build did NOT complete successfully

I tried the -host_force_python=PY2 but maybe I added it to the wrong place which is why it didn't work. I also don't remember having to do this the first time but I might be wrong. I would also like to add that I cloned my repository on my clean build. Perhaps there are some config files that need to be cleaned that I am not aware of?

I've also tried a clean clone of mediapipe using the handtracking example and I get the same error.

Does anyone have any suggestions?

Thank you for your time, Alex

closed time in a month

Meldanen

issue commentgoogle/mediapipe

how to specify a video in detection

You are welcome!

Meldanen

comment created time in a month

issue closedgoogle/mediapipe

I wan to try hand detection on a video I have. Where I can specify the video path?

closed time in a month

Chandler-Shuyu-Xie

issue commentgoogle/mediapipe

how to specify a video in detection

GLOG_logtostderr=1 bazel-bin/mediapipe/examples/desktop/hand_tracking/hand_tracking_cpu \
  --calculator_graph_config_file=mediapipe/graphs/hand_tracking/hand_tracking_desktop_live.pbtxt \
  --input_video_path=/path/to/input/video \
  --output_video_path=/path/to/output/video \

See details at https://github.com/google/mediapipe/blob/252a5713c7996e14c9bcff2caf0b2e1afca1e6f7/mediapipe/examples/desktop/demo_run_graph_main.cc#L36

Chandler-Shuyu-Xie

comment created time in a month

issue closedgoogle/mediapipe

Unable to build Android AAR

Environment

Ubuntu 20.04 64-bit
Bazel 3.2.0
GCC 9.3.0
OpenJDK 1.8.0_252
ANDROID_SDK, ANDROID_NDK_HOME and JAVA_HOME are set correctly
MediaPipe current master branch (commit 67bd8a2bf04072b853142e7739466dea218ae007)

Noteworthy

I am able to build & run the Android examples that ship with MediaPipe successfully.

Issue

I have a custom library/application based on the MediaPipe hand tracking example. I'm able to build this application either as a standalone executable or as a shared library on an Ubuntu machine and run it on the desktop without problems. Next my goal is to build an Android AAR to get the library functionality into an Android project. To do so, I've added the mediapipe_aar section to the bazel BUILD file as instructed by the official documentation:

package(default_visibility = [
    "//visibility:public",
])

cc_binary(
    name = "libmediapipe.so",
    srcs = ["hand_tracker.cpp",
            "hand_tracker.hpp", 
            "hand_tracker_impl.cpp", 
            "hand_tracker_impl.hpp", 
            "landmarks.hpp"],
    deps = [
        "@mediapipe//mediapipe/framework:calculator_framework",
        "@mediapipe//mediapipe/framework/formats:image_frame",
        "@mediapipe//mediapipe/framework/formats:image_frame_opencv",
        "@mediapipe//mediapipe/framework/formats:landmark_cc_proto",
        "@mediapipe//mediapipe/framework/port:commandlineflags",
        "@mediapipe//mediapipe/framework/port:file_helpers",
        "@mediapipe//mediapipe/framework/port:opencv_highgui",
        "@mediapipe//mediapipe/framework/port:opencv_imgproc",
        "@mediapipe//mediapipe/framework/port:opencv_video",
        "@mediapipe//mediapipe/framework/port:parse_text_proto",
        "@mediapipe//mediapipe/framework/port:status",
        "@mediapipe//mediapipe/graphs/hand_tracking:desktop_tflite_calculators",
    ],
    linkshared=True,
)

cc_binary(
    name = "libmediapipe_gpu.so",
    srcs = ["hand_tracker.cpp",
            "hand_tracker.hpp", 
            "hand_tracker_impl.cpp", 
            "hand_tracker_impl.hpp", 
            "landmarks.hpp"],
    local_defines = ["MEDIAPIPE_GPU"],
    deps = [
        "@mediapipe//mediapipe/framework:calculator_framework",
        "@mediapipe//mediapipe/framework/formats:image_frame",
        "@mediapipe//mediapipe/framework/formats:image_frame_opencv",
        "@mediapipe//mediapipe/framework/formats:landmark_cc_proto",
        "@mediapipe//mediapipe/framework/port:file_helpers",
        "@mediapipe//mediapipe/framework/port:opencv_highgui",
        "@mediapipe//mediapipe/framework/port:opencv_imgproc",
        "@mediapipe//mediapipe/framework/port:opencv_video",
        "@mediapipe//mediapipe/framework/port:parse_text_proto",
        "@mediapipe//mediapipe/framework/port:status",
        "@mediapipe//mediapipe/gpu:gl_calculator_helper",
        "@mediapipe//mediapipe/gpu:gpu_buffer",
        "@mediapipe//mediapipe/gpu:gpu_shared_data_internal",
        "@mediapipe//mediapipe/graphs/hand_tracking:mobile_calculators",
    ],
    linkshared=True,
)

load("@mediapipe//mediapipe/java/com/google/mediapipe:mediapipe_aar.bzl", "mediapipe_aar")

mediapipe_aar(
    name = "mp_hand_tracking_aar",
    calculators = ["@mediapipe//mediapipe/graphs/hand_tracking:mobile_calculators"],
)

However, the build process yields a visibility error upon trying to compile to an Android AAR:

joel@joel-ubuntu:~/projects/mediapipe_latest/my_custom_project/library$ bazel build -c opt --host_crosstool_top=@bazel_tools//tools/cpp:toolchain --fat_apk_cpu=arm64-v8a,armeabi-v7a :mp_hand_tracking_aar
ERROR: /home/joel/projects/mediapipe_latest/my_custom_project/library/BUILD:74:14: in genrule rule //my_custom_project/library:rasterization_proto_proto_java_src_generator: target '//mediapipe/framework/formats/annotation:protos_src' is not visible from target '//my_custom_project/library:rasterization_proto_proto_java_src_generator'. Check the visibility declaration of the former target if you think the dependency is legitimate
ERROR: Analysis of target '//my_custom_project/library:mp_hand_tracking_aar' failed; build aborted: Analysis failed
INFO: Elapsed time: 0.170s
INFO: 0 processes.
FAILED: Build did NOT complete successfully (0 packages loaded, 0 targets configured)
joel@joel-ubuntu:~/projects/mediapipe_latest/my_custom_project/library$

What exactly causes this issue and what's the intended way to make this build work?

closed time in a month

Tectu

issue commentgoogle/mediapipe

Unable to build Android AAR

We don't plan to expose the proto source code because that's some implementation details we may change later. So, we set the protos_src file_group to be only visible within the MediaPipe repo. Moreover, MediaPipe AAR rule was designed for packaging existing MediaPipe solutions such as face detection, hand tracking, and etc. So, we was expecting the users to use the aar rule within the MediaPipe repo.

For your case, you can change https://github.com/google/mediapipe/blob/master/mediapipe/framework/formats/annotation/BUILD#L43 to be visibility = ["//visibility:public"], to solve this issue.

Tectu

comment created time in a month

issue closedgoogle/mediapipe

run Custom tflite Model fails in TfLiteInferenceCalculator

train model by deeplapV3+ for tf2.0 and optimizar it by tensorflow_model_optimization 0.3.0 debug it and got log below:

2020-07-03 16:01:52.416 18260-18293/? D/mediapipeDebug: Following operations are not supported by GPU delegate: BATCH_TO_SPACE_ND: Operation is not supported. EXP: Operation is not supported. REDUCE_MAX: Operation is not supported. SPACE_TO_BATCH_ND: Operation is not supported. SUM: Operation is not supported. 148 operations will run on the GPU, and the remaining 90 operations will run on the CPU. 2020-07-03 16:01:52.422 18260-18293/? D/mediapipeDebug: TfLiteGpuDelegate Prepare: Only identical batch dimension is supported 2020-07-03 16:01:52.422 18260-18293/? D/mediapipeDebug: Node number 238 (TfLiteGpuDelegate) failed to prepare. 2020-07-03 16:01:52.422 18260-18293/? D/mediapipeDebug: external/org_tensorflow/tensorflow/lite/kernels/conv.cc:345 bias->type != input_type (10 != 1) 2020-07-03 16:01:52.422 18260-18293/? D/mediapipeDebug: Node number 2 (CONV_2D) failed to prepare

could anyone help me?

closed time in a month

forbywang

issue commentgoogle/mediapipe

run Custom tflite Model fails in TfLiteInferenceCalculator

The error message is BATCH_TO_SPACE_ND: Operation is not supported. EXP: Operation is not supported. REDUCE_MAX: Operation is not supported. SPACE_TO_BATCH_ND: Operation is not supported.. So, those ops are not supported by TFLite gpu. You need to either remove those ops from the graph or define those custom ops like we did here: https://github.com/google/mediapipe/tree/master/mediapipe/util/tflite/operations.

forbywang

comment created time in a month

issue commentgoogle/mediapipe

Unable to build example

The error message is ImportError: No module named numpy Is numpy installed?. Have you installed numpy via pip3?

Chandler-Shuyu-Xie

comment created time in a month

issue closedgoogle/mediapipe

Errors building on windows

System information

OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Windows 10
Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device:
TensorFlow installed from (source or binary):
TensorFlow version: 2.2.0, installed by mediapipe
Python version: 3.8.3
Installed using virtualenv? pip? conda?: conda
Bazel version (if compiling from source): 3.3.0
GCC/Compiler version (if compiling from source):
CUDA/cuDNN version:
GPU model and memory:

Describe the problem

Trying to build a mediapipe target:

bazel run mediapipe/examples/desktop/multi_hand_tracking:multi_hand_tracking_cpu

I get:

Loading: 
Loading: 0 packages loaded  
WARNING: Download from https://mirror.bazel.build/github.com/tensorflow/tensorflow/archive/7c09d15f9fcc14343343c247ebf5b8e0afe3e4aa.tar.gz failed: class com.google.devtools.build.lib.bazel.repository.downloader.UnrecoverableHttpException GET returned 404 Not Found     
WARNING: Download from http://mirror.tensorflow.org/github.com/bazelbuild/rules_closure/archive/cf1e44edb908e9616030cc83d085989b8e6cd6df.tar.gz failed: class com.google.devtools.build.lib.bazel.repository.downloader.UnrecoverableHttpException GET returned 404 Not Found   

Analyzing: target //mediapipe/examples/desktop/multi_hand_tracking:multi_hand_tracking_cpu (0 packages loaded, 0 targets configured)  
INFO: Repository local_execution_config_python instantiated at:  
no stack (--record_rule_instantiation_callstack not enabled)     
Repository rule local_python_configure defined at:        
C:/users/amit/_bazel_amit/j6xr7hh5/external/org_tensorflow/third_party/py/python_configure.bzl:275:41: in                                                                                                                                  
ERROR: An error occurred during the fetch of repository 'local_execution_config_python':          Traceback (most recent call last):                     
File "C:/users/amit/_bazel_amit/j6xr7hh5/external/org_tensorflow/third_party/py/python_configure.bzl", 
line 208                                                                                              
get_python_bin(repository_ctx)                                                              
File "C:/users/amit/_bazel_amit/j6xr7hh5/external/org_tensorflow/third_party/remote_config/common.bzl", line 44, in 
get_python_bin                                                                                                    
which(repository_ctx, "python")           
File "C:/users/amit/_bazel_amit/j6xr7hh5/external/org_tensorflow/third_party/remote_config/common.bzl", line 27, in which  execute(repository_ctx, <1 more arguments>)            
File "C:/users/amit/_bazel_amit/j6xr7hh5/external/org_tensorflow/third_party/remote_config/common.bzl", line 208, in execute 
                                 fail(<1 more arguments>)`

Provide the exact sequence of commands / steps that you executed before running into the problem

conda create --name=test python=3
conda activate test
git clone https://github.com/google/mediapipe.git
cd mediapipe
bazel run mediapipe/examples/desktop/multi_hand_tracking:multi_hand_tracking_cpu

closed time in a month

AmitMY

issue commentgoogle/mediapipe

Errors building on windows

In https://google.github.io/mediapipe/getting_started/install#installing-on-windows, we mention that adding --action_env PYTHON_BIN_PATH="C:/path/to/python.exe" is needed for Windows.

AmitMY

comment created time in a month

issue commentgoogle/mediapipe

C++ compilation error when building mp_multi_hand_aar

Can you show the full log? From you comment, we only see that it's caused by mediapipe/calculators/tflite:tflite_converter_calculator.

I guess the slowness of the build process is caused by downloading dependency libraries such as opencv android sdk and tensorflow repo?

AlexYiningLiu

comment created time in a month

issue closedgoogle/mediapipe

ERRORS when I run Desktop Helloworld

I completed the requirement configuration according to the instructions of Installing on Debian and Ubuntu but When running the last command:

$ export GLOG_logtostderr=1
# if you are running on Linux desktop with CPU only
$ bazel run --define MEDIAPIPE_DISABLE_GPU=1 \
    mediapipe/examples/desktop/hello_world:hello_world

Encountered the following error:

Starting local Bazel server and connecting to it...
INFO: SHA256 (https://github.com/bazelbuild/rules_cc/archive/master.zip) = a07fc8b90c22e42962142504438a84b31567976adcd9ee1a4c9f75d2e3396f8d
DEBUG: Rule 'rules_cc' indicated that a canonical reproducible form can be obtained by modifying arguments sha256 = "a07fc8b90c22e42962142504438a84b31567976adcd9ee1a4c9f75d2e3396f8d"
DEBUG: Repository rules_cc instantiated at:
  no stack (--record_rule_instantiation_callstack not enabled)
Repository rule http_archive defined at:
  /home/zcw/.cache/bazel/_bazel_zcw/43c4a13803b0469a6bd3b3fd6f82c078/external/bazel_tools/tools/build_defs/repo/http.bzl:336:16: in <toplevel>
INFO: Repository local_execution_config_python instantiated at:
  no stack (--record_rule_instantiation_callstack not enabled)
Repository rule local_python_configure defined at:
  /home/zcw/.cache/bazel/_bazel_zcw/43c4a13803b0469a6bd3b3fd6f82c078/external/org_tensorflow/third_party/py/python_configure.bzl:275:26: in <toplevel>
ERROR: An error occurred during the fetch of repository 'local_execution_config_python':
   Traceback (most recent call last):
	File "/home/zcw/.cache/bazel/_bazel_zcw/43c4a13803b0469a6bd3b3fd6f82c078/external/org_tensorflow/third_party/py/python_configure.bzl", line 208
		get_python_bin(repository_ctx)
	File "/home/zcw/.cache/bazel/_bazel_zcw/43c4a13803b0469a6bd3b3fd6f82c078/external/org_tensorflow/third_party/remote_config/common.bzl", line 44, in get_python_bin
		which(repository_ctx, "python")
	File "/home/zcw/.cache/bazel/_bazel_zcw/43c4a13803b0469a6bd3b3fd6f82c078/external/org_tensorflow/third_party/remote_config/common.bzl", line 29, in which
		execute(repository_ctx, <1 more arguments>)
	File "/home/zcw/.cache/bazel/_bazel_zcw/43c4a13803b0469a6bd3b3fd6f82c078/external/org_tensorflow/third_party/remote_config/common.bzl", line 208, in execute
		fail(<1 more arguments>)
Repository command failed

ERROR: Analysis of target '//mediapipe/examples/desktop/hello_world:hello_world' failed; build aborted: Traceback (most recent call last):
	File "/home/zcw/.cache/bazel/_bazel_zcw/43c4a13803b0469a6bd3b3fd6f82c078/external/org_tensorflow/third_party/py/python_configure.bzl", line 208
		get_python_bin(repository_ctx)
	File "/home/zcw/.cache/bazel/_bazel_zcw/43c4a13803b0469a6bd3b3fd6f82c078/external/org_tensorflow/third_party/remote_config/common.bzl", line 44, in get_python_bin
		which(repository_ctx, "python")
	File "/home/zcw/.cache/bazel/_bazel_zcw/43c4a13803b0469a6bd3b3fd6f82c078/external/org_tensorflow/third_party/remote_config/common.bzl", line 29, in which
		execute(repository_ctx, <1 more arguments>)
	File "/home/zcw/.cache/bazel/_bazel_zcw/43c4a13803b0469a6bd3b3fd6f82c078/external/org_tensorflow/third_party/remote_config/common.bzl", line 208, in execute
		fail(<1 more arguments>)
Repository command failed

INFO: Elapsed time: 13.957s
INFO: 0 processes.
FAILED: Build did NOT complete successfully (9 packages loaded, 11 targets configured)
FAILED: Build did NOT complete successfully (9 packages loaded, 11 targets configured)
    currently loading: @bazel_tools//tools/jdk

Before asking the question, I searched for related issues, but still failed to solve it, so I hope to seek some help :)

closed time in a month

ZZZhaowendao

issue commentgoogle/mediapipe

ERRORS when I run Desktop Helloworld

You successfully built bazel-bin/mediapipe/examples/desktop/hello_world/hello_world. Then, you need to run it ;)

Try the following:

$ export GLOG_logtostderr=1
$ bazel-bin/mediapipe/examples/desktop/hello_world/hello_world

ZZZhaowendao

comment created time in a month

issue commentgoogle/mediapipe

ERRORS when I run Desktop Helloworld

The issue is Bazel doesn't find the local python binary. You can run $ which python to find where the python binary is and then add --action_env PYTHON_BIN_PATH=<path to python binary> to the Bazel command like the following

bazel build -c opt --define MEDIAPIPE_DISABLE_GPU=1 --action_env PYTHON_BIN_PATH="/path/to/python" mediapipe/examples/desktop/hello_world

ZZZhaowendao

comment created time in a month

issue closedgoogle/mediapipe

Hello,

i'm trying to build the hand_tracking_gpu demo on a jetson nano board, but i'm getting the error Expected even number of arguments , building without cuda gives no error

rui@rui-jetson:~/pen/mediapipe$ bazel build -c opt --config=cuda --spawn_strategy=local     --define no_aws_support=true --copt -DMESA_EGL_NO_X11_HEADERS  mediapipe/examples/desktop/hand_tracking:hand_tracking_gpu
INFO: Repository local_config_cuda instantiated at:
  no stack (--record_rule_instantiation_callstack not enabled)
Repository rule cuda_configure defined at:
  /home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/gpus/cuda_configure.bzl:1235:33: in <toplevel>
ERROR: An error occurred during the fetch of repository 'local_config_cuda':
   Traceback (most recent call last):
        File "/home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/gpus/cuda_configure.bzl", line 1205
                _create_local_cuda_repository(<1 more arguments>)
        File "/home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/gpus/cuda_configure.bzl", line 931, in _create_local_cuda_repository
                _find_libs(repository_ctx, <2 more arguments>)
        File "/home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/gpus/cuda_configure.bzl", line 578, in _find_libs
                _check_cuda_libs(repository_ctx, <2 more arguments>)
        File "/home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/gpus/cuda_configure.bzl", line 480, in _check_cuda_libs
                execute(repository_ctx, <1 more arguments>)
        File "/home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/remote_config/common.bzl", line 208, in execute
                fail(<1 more arguments>)
Repository command failed
Expected even number of arguments
DEBUG: Rule 'rules_cc' indicated that a canonical reproducible form can be obtained by modifying arguments sha256 = "a07fc8b90c22e42962142504438a84b31567976adcd9ee1a4c9f75d2e3396f8d"
DEBUG: Repository rules_cc instantiated at:
  no stack (--record_rule_instantiation_callstack not enabled)
Repository rule http_archive defined at:
  /home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/bazel_tools/tools/build_defs/repo/http.bzl:336:31: in <toplevel>
ERROR: /home/rui/pen/mediapipe/mediapipe/examples/desktop/hand_tracking/BUILD:36:10: //mediapipe/examples/desktop/hand_tracking:hand_tracking_gpu depends on @local_config_cuda//crosstool:toolchain in repository @local_config_cuda which failed to fetch. no such package '@local_config_cuda//crosstool': Traceback (most recent call last):
        File "/home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/gpus/cuda_configure.bzl", line 1205
                _create_local_cuda_repository(<1 more arguments>)
        File "/home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/gpus/cuda_configure.bzl", line 931, in _create_local_cuda_repository
                _find_libs(repository_ctx, <2 more arguments>)
        File "/home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/gpus/cuda_configure.bzl", line 578, in _find_libs
                _check_cuda_libs(repository_ctx, <2 more arguments>)
        File "/home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/gpus/cuda_configure.bzl", line 480, in _check_cuda_libs
                execute(repository_ctx, <1 more arguments>)
        File "/home/rui/pen/bazel_cache/_bazel_rui/2863b3cfed4551e7bbfa622d9a9a5606/external/org_tensorflow/third_party/remote_config/common.bzl", line 208, in execute
                fail(<1 more arguments>)
Repository command failed
Expected even number of arguments
ERROR: Analysis of target '//mediapipe/examples/desktop/hand_tracking:hand_tracking_gpu' failed; build aborted: Analysis failed
INFO: Elapsed time: 0,747s
INFO: 0 processes.
FAILED: Build did NOT complete successfully (0 packages loaded, 0 targets configured)
    Fetching @local_config_cc; fetching
rui@rui-jetson:~/pen/mediapipe$

closed time in a month

ruisebastiao

issue commentgoogle/mediapipe

Thanks for sharing your solution!

Just to clarify, for hand tracking gpu, you don't need to config TensorFlow CUDA for MediaPipe. Hand tracking gpu example is doing inference with TFLite and its underlying OpenGL backend. The only MediaPipe example that needs CUDA is mediapipe/examples/desktop/object_detection:object_detection_tensorflow. We are still investigate how to make that work on Jetson devices.

ruisebastiao

comment created time in a month

issue commentgoogle/mediapipe

adb can be installed separately without Android Studio. So, I am not sure if the issue is related to the Android Studio version. The best way is to use the adb install command to install some other apps from the terminal. It this works, then try your robotcontrol app.

Meldanen

comment created time in a month

issue commentgoogle/mediapipe

https://github.com/tensorflow/tensorflow/issues/1171#issuecomment-244819545 might be helpful.

ruisebastiao

comment created time in a month

issue commentgoogle/mediapipe

ValidateRequiredSidePackets failed error when modifying the face tracking example to shoe detection.

The error is adb: unknown host service, which means that the apk was successfully built but adb is not able to install the apk. I suggest verifying your adb setting and make sure you can install the other apks as the first step. You also can build the apk by Bazel and then ask adb to install it separately.

bazel build  -c opt   --config android_arm64 mediapipe/examples/android/src/java/com/google/mediapipe/apps/robotcontrol:mainapp

adb install -r <path to your apk>

Meldanen

comment created time in a month

issue commentgoogle/mediapipe

Hi, unlike the face and hand examples, object detection 3d example requires extra input side packets. You need to send four input side packets into the graph in MainActivity.java. See https://github.com/google/mediapipe/blob/master/mediapipe/examples/android/src/java/com/google/mediapipe/apps/objectdetection3d/MainActivity.java#L50

davitgoder

comment created time in a month

issue closedgoogle/mediapipe

ValidateRequiredSidePackets failed error when modifying the face tracking example to shoe detection.

I've followed these instructions to get assets and .aar file for shoe detection, then I imported them as shown here

object_detection_3d_sneakers.tflite
objectdetection3d.binarypb
ssdlite_object_detection_labelmap.txt into assets folder and and
mp_object_detection_3d.aar

into libs folder

I used the most resembling example project in my opinion face detection and swapped respective files as mentioned above.

the build is successful, but after running it on the android device I get an error:

2020-06-29 15:43:44.002 1114-1312/com.example.myfacedetectionapp E/native: E0629 15:43:44.002478 1312 graph.cc:415] INVALID_ARGUMENT: ValidateRequiredSidePackets failed to validate: ; Side packet "box_asset_name" is required but was not provided. ; Side packet "box_texture" is required but was not provided. ; Side packet "obj_asset_name" is required but was not provided. ; Side packet "obj_texture" is required but was not provided. 2020-06-29 15:43:44.006 1114-1312/com.example.myfacedetectionapp E/AndroidRuntime: FATAL EXCEPTION: ExternalTextureConverter Process: com.example.myfacedetectionapp, PID: 1114 com.google.mediapipe.framework.MediaPipeException: invalid argument: ValidateRequiredSidePackets failed to validate: ; Side packet "box_asset_name" is required but was not provided. ; Side packet "box_texture" is required but was not provided. ; Side packet "obj_asset_name" is required but was not provided. ; Side packet "obj_texture" is required but was not provided. at com.google.mediapipe.framework.Graph.nativeStartRunningGraph(Native Method) at com.google.mediapipe.framework.Graph.startRunningGraph(Graph.java:289) at com.google.mediapipe.components.FrameProcessor.startGraph(FrameProcessor.java:372) at com.google.mediapipe.components.FrameProcessor.maybeAcceptNewFrame(FrameProcessor.java:296) at com.google.mediapipe.components.FrameProcessor.onNewFrame(FrameProcessor.java:311) at com.google.mediapipe.components.ExternalTextureConverter$RenderThread.renderNext(ExternalTextureConverter.java:275) at com.google.mediapipe.components.ExternalTextureConverter$RenderThread.lambda$onFrameAvailable$0$ExternalTextureConverter$RenderThread(ExternalTextureConverter.java:221) at com.google.mediapipe.components.-$$Lambda$ExternalTextureConverter$RenderThread$Y1vV_XyLsWZ0ebOvq-iwjQ0H3Sw.run(Unknown Source:4) at android.os.Handler.handleCallback(Handler.java:883) at android.os.Handler.dispatchMessage(Handler.java:100) at android.os.Looper.loop(Looper.java:214) at com.google.mediapipe.glutil.GlThread.run(GlThread.java:134)

closed time in a month

davitgoder

issue commentgoogle/mediapipe