Fixed the PyPi links (per #649 PR)

This release adds a variety of new options to the PipelineManager. Now, it's possible to change whitebalance, autowhitebalance mode, autofocus mode, or manual focus. It's also possible to trigger autofocus or still capture

Features

• Add colorOrder to createColorCamera in pipeline manager in SDK.

In this small release, we fixed an issue with the missing numpy import statement in PreviewManager

Features

• In BlobManager, added progressFunc parameter to init method allowing to override default show mechanism and zooType to getBlob method, so that the download from our model zoo is now possible
• Color map used by PreviewManager will use black ([0, 0, 0]) as lowest value, to show a better visual separation between valid and invalid points on the depth preview
• Added createBlankFrame to util methods, allowing to create a blank frame with specified width, height and optionally color

Bugs

• Use numpy 1.21 for Python >3.7
• Removed flipDetection from NNetManager as rectified streams are now flipped on device
• PreviewManager now calls fpsHandler automatically on each received packet

Misc

• Disabled stream encoding in low-bandwidth mode for rectified and disparity streams to prevent OOM issue when too many streams are requested at once

In this release, we added a couple of minor improvements:

• Added poeQuality init parameter for PipelineManager object - determines the PoE encoding quality. Decreasing this value can reduce the display latency and improve overall FPS output for OAK-1-POE or low-throughput networks (value range: 0-100)
• Fixed queues closing and variables initialization, so that multiple inits / destroys with SDK objects is possible (e.g. creating 3 instances of PipelineManager will work correctly)
• Limited the maxDataSize parameter for input queues - for image inputs, set max data size to equal nn input size, whereas for config inputs (depth config, camera configs) it's set to 1KB. This helps in limiting the memory used on the device and allows more configurations to be run
• LRC (left-right check) is now configurable during the runtime using PipelineManager.updateDepthConfig, setting lrc param to True if you want to enable it
• Introduced encQuality to PipelineManager.createEncoder method that allows to specify the encoding quality (value range: 0-100)
• Added createWindows init param to PreviewManager (default True) that determines if PreviewManager will show OpenCV preview windows. If you'd like to use other preview methods (e.g. send frames over a network or store them locally), you can set this param to False and use the callback param to provide a processing function
• Added subpixel support for PreviewDecoder so that depth output will contain a valid disparity output if subpixel mode is enabled in stereo depth node
• Added resizeLetterbox to utils that allows resizing the image to a specified size, but preserve its aspect ratio, adding black borders when necessary

In this releaase, we added maxTicks parameter to FPSHandler, allowing to specify the amount of ticks taken into consideration while calculating FPS. More details can be found in the PR

Also, with this PR, the package requirements got loosened - we now require just major versions of OpenCV and DepthAI libraries and same-or-greater indicators to other dependencies. We hope that with this change, the SDK will cooperate better with other package configurations

Important - blobconverter update

During this weekend, we experienced issues with the current blobconverter infrastructure. After making it work again, the HTTP communication was disabled and blobconverter v1.2.2 was released that uses the updated HTTPS URL. All previous versions of blobconverter will try to connect via HTTP and will throw an error To mitigate this issue, we strongly recommend installing the latest version of both DepthAI SDK and blobconverter in particular.

For those who would not like to update or just can't, please use the code below to point blobconverter to a different URL

```python import blobconverter

blobconverter.set_defaults(url="https://blobconverter.luxonis.com") ```

In this release, we added a capability to set the camera settings.

Three new functions were added to PipelineManager: - updateColorCamConfig - to update color camera config - updateLeftCamConfig - to update left camera config - updateRightCamConfig - to update right camera config

Each of these functions can modify the following camera attributes: - exposure - sensitivity - saturation - contrast - brightness - sharpness

To know more about these attributes, check our documentation

You can test these options using our demo script: - Control using a keyboard by running $ python3 depthai_demo.py --cameraControlls - Set specific camera configuration from start using command-line arguments (all flags listed here) $ python3 depthai_demo.py --cameraBrightness 5

DepthAI SDK is a Python package, containing convenience classes and functions that help in most common tasks while using DepthAI API.

Package documentation - https://docs.luxonis.com/projects/sdk/en/latest/ PyPI page - https://pypi.org/project/depthai-sdk/

To install the package, please run the following command

pip install depthai-sdk

To see this package in action, you can check the getting started guide or the following experiments:

• gen2-people-counter: https://github.com/luxonis/depthai-experiments/tree/master/gen2-people-counter
• gen2-human-pose: https://github.com/luxonis/depthai-experiments/tree/master/gen2-human-pose
• gen2-road-segmentation: https://github.com/luxonis/depthai-experiments/tree/master/gen2-road-segmentation

To know more about the DepthAI itself, visit our documentation or our webpage