Mediapipe

* Better pose landmark accuracy with the updated pose detection and pose landmark model
* Added a `lite` and a `heavy` version of the pose landmark model, in addition to the existing `full` version (also updated and improved)
* Added a `model_complexity` config option in [MediaPipe Pose](https://solutions.mediapipe.dev/pose#model_complexity) and [MediaPipe Holistic](https://solutions.mediapipe.dev/holistic#model_complexity) Solution APIs to select across the 3 model versions
* Removed the `upper_body_only` option in MediaPipe Pose and MediaPipe Holistic, as the standard model now already handles upper-body-only use cases well

* Move up minimum bazel version to [3.7.2](https://github.com/google/mediapipe/blob/7c331ad58b2cca0dca468e342768900041d65adc/WORKSPACE#L17), and encourage the use of [Bazelisk](https://docs.bazel.build/versions/master/install-bazelisk.html) in [Installation](https://google.github.io/mediapipe/getting_started/install.html).
* Move up [TensorFlow dependency](https://github.com/google/mediapipe/blob/7c331ad58b2cca0dca468e342768900041d65adc/WORKSPACE#L369).
* MediaPipe Objectron [desktop/C++ example](https://solutions.mediapipe.dev/objectron#desktop).

* [MediaPipe Pose](https://solutions.mediapipe.dev/pose)
* New pose landmark model with Z coordinates ([BlazePose GHUM 3D](https://solutions.mediapipe.dev/pose#pose-landmark-model-blazepose-ghum-3d)), generated from synthetic data
* Added tutorials and Colabs for [Pose Classification](https://solutions.mediapipe.dev/pose_classification)
* [MediaPipe Objectron](https://solutions.mediapipe.dev/objectron)
* Added Python Solution API
* [MediaPipe Face Detection](https://solutions.mediapipe.dev/face_detection)
* Added Python Solution API and JavaScript Solution API
* [MediaPipe Face Mesh](https://solutions.mediapipe.dev/face_mesh)
* [Face Geometry Module](https://solutions.mediapipe.dev/face_mesh#face-geometry-module) ([code](https://github.com/google/mediapipe/tree/master/mediapipe/modules/face_geometry)) now supports [face-detection input](https://github.com/google/mediapipe/blob/master/mediapipe/modules/face_geometry/face_geometry_from_detection.pbtxt) (to generate a face geometry from a detection), in addition to the existing support for [face-landmark input](https://github.com/google/mediapipe/blob/master/mediapipe/modules/face_geometry/face_geometry_from_landmarks.pbtxt)
* [Image](https://github.com/google/mediapipe/blob/master/mediapipe/framework/formats/image.h): a new multi-backend image data container:
* Covering both CPU and GPU storage
* Currently used in [ImageToTensorCalculator](https://github.com/google/mediapipe/blob/master/mediapipe/calculators/tensor/image_to_tensor_calculator.cc) and [Face Detection Module](https://github.com/google/mediapipe/tree/master/mediapipe/modules/face_detection)
* Companion utility [ToImageCalculator](https://github.com/google/mediapipe/blob/master/mediapipe/calculators/util/to_image_calculator.cc) and [FromImageCalculator](https://github.com/google/mediapipe/blob/master/mediapipe/calculators/util/from_image_calculator.cc)

* Released [experimental new framework APIs](https://github.com/google/mediapipe/tree/master/mediapipe/framework/api2/README.md)
* Updated [MediaPipe Objectron](https://solutions.mediapipe.dev/objectron) to support simultaneous detection of multiple objects
* Bug fixes
* Fixed a crash in the [MediaPipe Iris](https://solutions.mediapipe.dev/iris) iOS example app
* Fixed memory leak in Android example apps (in [ExternalTextureConverter](https://github.com/google/mediapipe/blob/master/mediapipe/java/com/google/mediapipe/components/ExternalTextureConverter.java))
* Fixed a file read mode issue in [pose_landmark_model_loader.pbtxt](https://github.com/google/mediapipe/blob/master/mediapipe/modules/pose_landmark/pose_landmark_model_loader.pbtxt#L51).

* [MediaPipe Holistic](https://solutions.mediapipe.dev/holistic) for real-time simultaneous tracking of human pose, face and hand landmarks
* [Google AI blog post](https://mediapipe.page.link/holistic_ai_blog_post)
* [MediaPipe in Python](https://docs.mediapipe.dev/getting_started/python)
* Ready-to-use Python Solution APIs for MediaPipe Face Mesh, Hands, Pose and Holistic
* Try it in [MediaPipe on Google Colab](https://docs.mediapipe.dev/getting_started/python#mediapipe-on-google-colab)
* [MediaPipe in JavaScript](https://docs.mediapipe.dev/getting_started/javascript)
* Ready-to-use JavaScript Solution APIs for MediaPipe Face Mesh, Hands, Pose and Holistic
* Try it in [MediaPipe on CodePen](https://code.mediapipe.dev/codepen)
* Switch [build](https://github.com/google/mediapipe/blob/master/.bazelrc) from C++14 to C++17

* [MediaPipe in Python](https://google.github.io/mediapipe/#mediapipe-in-python)
* Released Python Solution API for [MediaPipe Face Mesh](https://solutions.mediapipe.dev/face_mesh#python) and [MediaPipe Hands](https://solutions.mediapipe.dev/hands#python)
* Updated Python Solution API for [MediaPipe Pose](https://solutions.mediapipe.dev/pose#python)
* Also released usage examples as [Google Colabs](https://google.github.io/mediapipe/#mediapipe-in-python)
* [MediaPipe Objectron](https://solutions.mediapipe.dev/objectron)
* Released a faster two-stage pipeline
* Added support for more object classes: shoe, chair, cup and camera
* Released [training dataset](https://github.com/google-research-datasets/Objectron/), (to be) announced in a [Google AI Blog post](https://mediapipe.page.link/objectron_dataset_ai_blog)
* New [Tensor](https://github.com/google/mediapipe/blob/master/mediapipe/framework/formats/tensor.h) class
* Released as a multi-dimensional tensor data container, supporting multiple backends like CPU, Metal buffer, GL buffer and GL texture 2D
* Added new Tensor-based pre-processing, inference and post-processing calculators in [mediapipe/calculators/tensor](https://github.com/google/mediapipe/tree/master/mediapipe/calculators/tensor), branched from existing calculators in [mediapipe/calculators/tflite](https://github.com/google/mediapipe/tree/master/mediapipe/calculators/tflite)
* Most of the (sub-)graph in [mediapipe/modules](https://github.com/google/mediapipe/tree/master/mediapipe/modules) have been updated to use Tensor and the associated calculators. The plan is to fully migrate all in the repo (and deprecate [mediapipe/calculators/tflite](https://github.com/google/mediapipe/tree/master/mediapipe/calculators/tflite) by end of 2020.
* [MediaPipe Hands](https://solutions.mediapipe.dev/hands)
* Refactored [graphs](https://github.com/google/mediapipe/tree/master/mediapipe/graphs/hand_tracking) to depend on the new [palm_detection](https://github.com/google/mediapipe/tree/master/mediapipe/modules/palm_detection) and [hand_landmark module](https://github.com/google/mediapipe/tree/master/mediapipe/modules/hand_landmark).
* Improved model speed for both palm detection and hand landmark.
* Extended the main hand tracking [example apps](http://solutions.mediapipe.dev/hands#example-apps) to support multiple hands, to replace the separate multi-hand tracking examples.
* [MediaPipe Face Detection](https://solutions.mediapipe.dev/face_detection)
* Refactored [graphs](https://github.com/google/mediapipe/tree/master/mediapipe/graphs/face_detection) to depend on the [face_detection module](https://github.com/google/mediapipe/tree/master/mediapipe/modules/face_detection).