MediaPipe v0.6.4 - Multi-hand tracking, BeginLoop/EndLoop calculators

MediaPipe v0.6.4 - Multi-hand tracking, BeginLoop/EndLoop calculators

• MediaPipe Multi-hand tracking example app
  • For Android, iOS GPU, desktop CPU/GPU
  • BeginLoop/EndLoop calculators for processing a collection of elements iteratively
• Updated dependencies
  • Support Bazel 1.1.0
  • Update TensorFlow dependency to the most recent version
  • Use ‘-std=c++14’ by default
  • Use Google absl’s flat_hash_set and flat_hash_map for better performance.
• Bug fixes
  • Fix a bug in OpenCVVideoDecoderCalculator for issue 204
  • Fix a bug in kinetics_dataset.py for issue 228
• Documentation
  • Multi-hands mobile & desktop
  • Updated list of public events for MediaPipe

MediaPipe v0.6.3 - YT8m inference example & Android AAR

v0.6.3

• MediaPipe Android development with Gradle
  • Building Android AAR using Bazel
  • Using MediaPipe built AAR in Android Studio
• YT8mil baseline model inference pipeline
• Bazel versions supported are v0.24.1 to v0.29.1
• Bug fixes
  • Android AAR for issue 88
• Documentation
  • Model card for YT8mil baseline model
  • MediaPipe Android development with Gradle
  • YT8M model inference

MediaPipe v0.6.2 - Desktop CPU & GPU examples

v0.6.2

• MediaPipe CPU & GPU graph runner for Desktop graphs with webcam
  • Mobile graphs example & doc
    - Face Detection
    - Mobile CPU
• Desktop graph examples & docs
  - Face detection
  - Desktop GPU
  - Desktop CPU
  - Object detection
  - Desktop CPU
  - Hand tracking
  - Desktop GPU
  - Desktop CPU
  - Hair segmentation
  - Desktop GPU
• Allows android to print log output from native code. Fixed issue 118
• Fix bug in YT8m feature extraction pipeline

MediaPipe v0.6.1 - YT8m Feature extractor & bug fixes

v0.6.1

• New Features
  • YT8m Feature extractor for Kaggle challenge for ICCV workshop
  • Deepmind Kinetics I3D pre-processing pipeline
• Bug fixes
  • OpenCVVideoDecoderCalculator for issue 57
• Documentation
  • Fix to Android studio setup instruction for issue 65
  • Fix to iOS setup bundle ID for issue 50
  • Fix to Android NDK location for issue 98

MediaPipe v0.6.0 - Obj-C APIs + Hand Examples

Major New Features with v0.6.0

• Subgraph API in C++
• Graph Execution API in Obj-C
• Updated previous mobile examples with iOS
  • Edge detection with GPU (Hello World! on iOS )
  • Object Detection with GPU
  • Face Detection with GPU
  • Face Detection with CPU
• 2 new mobile examples with Android & iOS
  • Hand detection with GPU
  • Hand tracking with GPU
• A new data preprocessing graph for preparing kinetics dataset
• Audio Calculators - decoding, resampling, spectrogram, mfcc
• Visualizer
  • Multi-tab support
  • Subgraph visualization
• Documentation
  • Added iOS section for edge detection, face detection, and object detection
  • Google AI blog post on hand tracking
  • Hand detection
  • Hand tracking
  • Hand model card
  • Subgraph visualization

MediaPipe v0.5.0

Major Features

• Calculator API in C++
• Graph Construction API in ProtoBuf
• OpenGL ES support for Calculator API
• Graph Execution API in C++
• Graph Execution API in Java (Android)
• 5 Desktop and Android examples (Graphs, Calculators, TF/TFlite models)
  • Edge detection on GPU for Android (Hello World! on Android)
  • Object Detection on CPU and GPU for Android
  • Face Detection on GPU for Android
  • Hair Segmentation on GPU for Android
  • Object Detection on Desktop (No desktop GPU support)
• Visualizer launched at https://viz.mediapipe.dev
• Dataset preprocessing using MediaSequence
• Set of 59 calculators for core framework, image, tensorflow, tflite, util, video
• Initial v0.5 documentation launched at https://docs.mediapipe.dev