View Code? Open in Web Editor
NEW
This project forked from prantoran/mediapipe
Cross-platform, customizable ML solutions for live and streaming media.
Home Page: https://mediapipe.dev
License: Apache License 2.0
Shell 0.21%
JavaScript 0.16%
C++ 80.33%
Python 3.71%
C 0.69%
Objective-C 1.17%
Java 3.87%
CSS 0.05%
Objective-C++ 1.36%
Makefile 0.01%
HTML 0.10%
Dockerfile 0.05%
Starlark 8.31%
mediapipe's Introduction
MediaPipe GSOC 2021: Web Video Effects App
- Built on-top of the MediaPipe project.
- The Emscripten mapping and the BUILD files for the project are in
hello-world
.
- Summary of the work done during GSOC 2021 is here.
- Set up MediaPipe using the official instructions.
- The WebGL rendering via WASM uses GPU, so we need GPU version of MediaPipe setup.
- The development environment for the project was Linux.
- From the root directory of the project, run the command in the terminal:
make build && make run
- The scripts that are executed are in
scripts/...
and Makefile
.
make build
compiles the WASM binaries:
bazel build -c opt //hello-world:hello-world-simple --config=wasm
make run
copies the required outputs and runs a NodeJS server to run the WASM binaries locally.
- Output files are copied from
bazel-out/wasm-opt/bin/hello-world/...
- WASM versions of face detection, selfie segmentation and face mesh are demonstated in the following web app.
mediapipe's People
Contributors