- rpi::ScalerCrops control support
- Hailo multi-model support
- Stereo preview example script
- Add PyavOutput and a new CircularOutput2
- Allow libav H264 encoder to use V4L2 hardware on VC4 platforms
- bbox-order argument for imx500_object_detection_demo script
- Fix ScalerCrops tests
- imx500: Update MAX_NUM_TENSORS and MAX_NUM_DIMENSIONS
- Fix V4L2 encoder not releasing requests
- Add to_tuple methods to libcamera Rectangle and Size types
- Add IMX500 support
- Fixed setup.py for the devices helpers
- Fixed handling of the ScalerCrops control in app_full.py
- Initial support for Hailo AI devices, including some examples.
- IMX708 helper class so that the sensor HDR mode can be set with Python.
- 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).
- 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.
- Configuration alignment fixed on Pi 5.
- Improve support for displays without alpha blending.
- Yolov5s TensorFlow example.
- MJPEG server example with 90/270 degree rotation.
- File descriptor leak fixed when closing DmaAllocator.
- EGL memory leak fied.
- Missing R16 mono raw format added for Pi 5.
- Minor LibavH264Encoder improvements.
- 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.
- FFmpeg thread queue size increased, to suppress warnings and improve behaviour.
- Quality argument is respected properly by start_encoder().
- A memory leak when using the new DmaAllocator has been fixed, allowing camera mode switches to work indefinitely again.
- capture_request() allows you to specify a "no earlier than" timestamp.
- A callback is available in case the FfmpegOutput terminates unexpectedly.
- Workaround for V4L2 encoders getting stuck when trying to flush output.
- Fix for PIL version 10.
- Configuration objects work better with the new SensorConfiguration e.g. "picam2.still_configuration.sensor.output_size = (2028, 1520)"
- Updates for the latest libcamera.
- Add picamera2.allocators package.
- Pi 5 support added. Includes platform detection, new raw formats etc.
- Libav encoders added. These will be used as the H264 and MJPEG encoder on Pi 5.
- Frame buffers are now cached to improve performance.
- Encoders should no longer drop frames when closed.
- FfmpegOutput should stop properly when there's an audio stream.
- Updates for the latest version of libcamera, including sensor configuration.
- Small bug that prevented the configuration of USB cams fixed.
- The search path for tuning files (after the libcamera re-organisation) has been fixed.
- The built-in configuration objects (picamera2.preview_configuration etc.) are updated to work better with raw streams now that the latest libcamera is more efficient when we use them.
- Support for running multiple encoders, either on the same or different streams.
- Pathlib objects now supported.
- The camera can be stopped more quickly.
- Fix problem passing modes from the sensor mode list to the configure methods.
- H264 encoder quality was having no effect on bitrate - fixed.
- ffmpegoutput - wait correctly for it to finish when we stop it.
- Amount of CMA memory required is reduced for many use cases.
- MJPEG server example added that uses the hardware MJPEG encoder.
- Example showing previews from two cameras in a single Qt app.
- H264 encoder can accept a frame time interval to be put in the SPS headers.
- H264 encoder should now advertise the correct level in streams for higher bitrates.
- Exif DateTime and DateTimeOriginal tags are now added.
- H264 encoder now supports a constant quality parameter.
- JpegEncoder chooses pixel format automatically from the stream format.
- Work around OpenGL import error on 64-bit OS Lite images.
- FFmpeg: use the audio_samplerate parameter correctly.
- camera_controls are converted to native Python types (tuples, not libcamera.Rectangle objects).
- Camera configuration will fail if an encoder is still running.
- Encoder interface tidied to make it easier to run a second encode "by hand" (example included).
- Picamera2 updated for the latest libcamera which changes colour space handling.
- Encoder outputs no longer close files that they didn't open.
- CircularOutput class now flushes the fifo correctly when outputs are closed.
- Bug fix when capturing several DNG files back to back.
- Support for cameras that have autofocus integrated properly with libcamera.
- New switch_mode_and_capture_request method.
- Fewer frame drops when recording videos.
- Fixes when closing a preview window with the mouse.
- Ability to control via the configuration "queue" parameter whether Picamera2 keeps hold of the last completed request or not. This means you may wait marginally longer for a capture in some use cases, but also that the system cannot give you the frame that arrived slightly before you requested it.
- The
Picamera2.start_encoderfunction prototype has been made very similar toPicamera2.start_recordingfor consistency. Most existing calls still work, but there are a few call patterns that may need updating. The new prototype is:start_encoder(self, encoder=None, output=None, pts=None, quality=Quality.MEDIUM) - The
Picamera2.waitfunction now requires an argument, which is the "job" that was returned to you when you made the asynchronous call, for example instead of
picam2.capture_file("test.jpg", wait=False)
metadata = picam2.wait()
you would use
job = picam2.capture_file("test.jpg", wait=False)
metadata = picam2.wait(job)
Please also refer to our updated Qt examples.
- Some bug fixes when starting and stopping encoders.
- Risk of video frame drops during transient busy periods reduced.
- The
Picamera2.global_camera_info()method will return information about all the attached cameras. - We have introduced the ability to control multiple Picamera2 objects (all opened for different cameras) within the same Python process. They behave independently and can each have their own preview window.
- There is now limited support for USB webcams that deliver MJPEG or YUYV streams. Images can be displayed by the QT (not QTGL) preview.
- Picamera2 objects have a title_fields property which can be set to a list of the metadata fields to display on the preview window title bar (for example `picam2.title_fields = ["ExposureTime", "AnalogueGain"]).
- Resources are freed more reliably when Picamera2 objects are closed.
- Add support for outputting timestamps when using the JPEG encoder for video.
- Fix a bug which prevented the mjpeg_server.py example from working. This has also been added to the automatic test suite so shouldn't get broken again!
- Ability to record uncompressed or raw video frames.
- Releasing of camera resources has been made more reliable, so once the Picamera2 object has been closed it can be re-opened more reliably (optionally by other processes).
- Logging has been revised to work better with applications already using Python logging.
- Some reporting of, and recovery from, certain error conditions has been improved.
- Large encoded video frames are now split automatically for streaming over UDP.
- Very minor changes for the latest version of libcamera. "libcamera.ColorSpace.Jpeg()" has become "libcamera.ColorSpace.Sycc()".
- All the preview implementations now support a "display transform", allowing the preview image to be horizontally and/or vertically flipped (whilst not affecting the underlying image). The Picamer2.start_preview method allows a libcamera transform to be passed.
- Added APIs to capture and copy buffers/arrays from multiple streams at once: capture_buffers, capture_arrays, switch_mode_and_capture_buffers, switch_mode_and_capture_arrays.
- Allow entries from the sensor_modes property to be used directly as the raw stream configuration.
- Support for version 2.0 tuning files, including a find_tuning_algo method to make them easier to use.
- Demo Qt applications have been moved out of the examples folder to apps. A new "app_full.py" exists which allows phots and videos to be recorded, and gives control through a GUI of various camera and image tuning parameters.
- Added a sensor_modes field to the Picamera2 object. This can be queried to find out exactly what raw camera modes are supported, giving details of the maximum framerate and the field of view.
- Installation through pip now avoids installing the Qt and OpenGL dependencies by default. If you want them, use "pip3 install picamera2[gui]" or just do "sudo apt install -y python3-qt5 python3-opengl" first.
- Fixed bug displaying overlays using DRM (pykms seems to have changed underneath us).
- There's been some refactoring which has changed the way asynchronous calls (with wait=False) work. You should now call picam2.wait() to obtain the result, and you can set the signal_function so that you can avoid calling it before it's finished. The previous fields like "async_result" have been removed.
- JpegEncoder defaults to producing YUV420 output now, though the constructor allows other colour subsampling modes to be chosen.
- Fix bug where framerates had to be integer values.
- Fix typo in align_configuration method.
- Allow in-place image manipulation (using the MappedArray class) even when the image rows have padding.
- High level API functions "start_and_capture_file/files", "start_and_record_video" are added for users who need to know less about the Picamera2 internals.
- A very trivial Metadata class is provided that can be used to wrap the metadata dictionaries, for those that prefer this style.
- A Controls class is provided which gives an object-like view to control lists. The class is able to check that the control names are valid.
- Configuration structures are provided so that the Picamera2 object can be configured using the "object" style. Three such instances are embedded in the Picamera2 object, namely preview_configuration, still_configuration and video_configuration.
- The start_recording() method accepts an optional configuration.
- The start_preview() method accepts True as an indication to try and autodetect the correct type of preview window. Note that there will be situations where it guesses incorrectly.
- The start() method can optionally be given config and show_preview parameters which will configure the camera and start the preview (if it isn't running already).
- Streams no longer have their widths optimally aligned by default. A separate align_configuration() method can be called to enforce this.
- The preview_configuration(), still_configuration() and video_configuration() methods (xxx_configuration) are renamed to create_xxx_configuration().
- The QPicamera2 widget will now display YUV420 images if the cv2 (OpenCV) module is installed.
- Methods that allow still captures to be saved to files have been updated to accept file-like objects, such as BytesIO.
- NOGUI=1 environmental variable to install without GUI dependencies
- The Picamera2's request_callback has been changed to post_callback. A new pre_callback has been added which runs before images copied for applications.
- Support for multiple outputs for the encoder
- The still_configuration defaults now specify the default display and encode streams as none. The resolutions are often so large that the images can be problematic and there are often better alternatives (e.g. specifying a lores stream and displaying that instead).
- The preview window implementations are changed to preserve image aspect ratios which they did not previously. The Qt variants also respect resize events.
- Everything installable via apt and pip.
- Many new examples.
- Revised API - please refer to README.md file and examples.