GitHub - kjxlstad/mediapipe: Cross-platform, customizable ML solutions for live and streaming media.

I had some troubles trying to configure this, but got it working in the end. Hopefully this can help others doing the similar changes in the future.

Cuda options and links to nvcc was added to .bazel
Path to opencv and ffmpeg was configured in workspace
Build options in /graphs/hand_tracking/ and /python/ was changed to point to the gpu graphs
Number of hands supported was changed in /graphs/hand_tracking/hand_tracking_desktop_live_gpu.pbtxt
The hand_landmark_tracking_gpu.pbtxt graph was modified for converting the images from python to gpu buffers
/python/solutions/hands.py was modified to utilize the gpu graphs
setup.py points to gpu graphs instead, flags for MESA_EGL_NO_X11_HEADERS and EGL_NO_X11 were also added
Paths were uptaed in opencv_linux.build and ffmpeg_linux.build were changed to point to the correct directory.

Name		Name	Last commit message	Last commit date
Latest commit History 99 Commits
docs		docs
mediapipe		mediapipe
third_party		third_party
.bazel		.bazel
.bazelrc		.bazelrc
.bazelversion		.bazelversion
.dockerignore		.dockerignore
.gitignore		.gitignore
BUILD.bazel		BUILD.bazel
CONTRIBUTING.md		CONTRIBUTING.md
Dockerfile		Dockerfile
LICENSE		LICENSE
MANIFEST.in		MANIFEST.in
README.md		README.md
WORKSPACE		WORKSPACE
build_android_examples.sh		build_android_examples.sh
build_desktop_examples.sh		build_desktop_examples.sh
build_ios_examples.sh		build_ios_examples.sh
requirements.txt		requirements.txt
setup.py		setup.py
setup_android_sdk_and_ndk.sh		setup_android_sdk_and_ndk.sh
setup_opencv.sh		setup_opencv.sh

Provide feedback