Picamera2

Changed

* Add to_tuple methods to libcamera Rectangle and Size types
* Add IMX500 support

Changed

* Fixed setup.py for the devices helpers
* Fixed handling of the ScalerCrops control in app_full.py

Added

* Initial support for Hailo AI devices, including some examples.
* IMX708 helper class so that the sensor HDR mode can be set with Python.

Changed

* Improved handling of timeouts when cameras stop responding, including a mechanism for a complete
reset if requests stop being returned.
* Platform checking more robust.
* Add missing flush parameter to captured_request() (for use with context manager).

Added

* Add an example showing how to forward images to other processes using zero-copy.
* Add a context manager method for capturing requests, e.g. `with picam2.captured_request() as r:`
* Encoders can skip frames, e.g. run at half the rate of the camera.

Changed

* Configuration alignment fixed on Pi 5.
* Improve support for displays without alpha blending.

Added

* Yolov5s TensorFlow example.
* MJPEG server example with 90/270 degree rotation.

Changed

* File descriptor leak fixed when closing DmaAllocator.
* EGL memory leak fied.
* Missing R16 mono raw format added for Pi 5.
* Minor LibavH264Encoder improvements.

Added

* Pi 5 users can now request 48-bit RGB outputs (where each R, G and B sample is 16-bit).
* Support for h.264 codec profiles.
* DNG files can be written to ioBytes objects.
* Persistent allocators added.
* Timeout support waiting for asynchronous requests to complete.

Changed

* FFmpeg thread queue size increased, to suppress warnings and improve behaviour.
* Quality argument is respected properly by start_encoder().