Azure Percept DK Container release notes

  • Article

Important

Retirement of Azure Percept DK:

Update 22 February 2023: A firmware update for the Percept DK Vision and Audio accessory components (also known as Vision and Audio SOM) is now available here, and will enable the accessory components to continue functioning beyond the retirement date.

The Azure Percept public preview will be evolving to support new edge device platforms and developer experiences. As part of this evolution the Azure Percept DK and Audio Accessory and associated supporting Azure services for the Percept DK will be retired March 30th, 2023.

Effective March 30th, 2023, the Azure Percept DK and Audio Accessory will no longer be supported by any Azure services including Azure Percept Studio, OS updates, containers updates, view web stream, and Custom Vision integration. Microsoft will no longer provide customer success support and any associated supporting services. For more information, please visit the Retirement Notice Blog Post.

This page provides information of changes and fixes for Azure Percept DK Container releases.

To download the container updates, go to Azure Percept Studio, select Devices from the left navigation pane, choose the specific device, and then select Vision and Speech tabs to initiate container downloads.

January (2201) Release

  • Security update for HostIPModule and ImageCapturingModule.

    Note

    HostIPModule and ImageCapturingModule are for vision AI-related experiences and will be deployed after the devkit connects to your instance of Azure Percept Studio. If these modules have been deployed to your devkit, complete the following steps to update them. If these two modules have not been deployed to your devkit, you can ignore the following steps and the latest modules will be deployed through Azure Percept Studio when needed.

    1. Find your devkit in your instance of Azure Percept Studio. In the summary of the device, select Open device in IoT Hub.
      1. In the IoT Hub UI of your devkit:
        1. Check the modules list to make sure HostIPModule and ImageCapturingModule have already been deployed.
        2. Select Set modules.
          1. In the deployment list, select the trashcan icon to remove HostIPModule and ImageCapturingModule.
          2. Select review + create, and then select create to finish the deployment.
      2. In the IoT Hub UI of your devkit, wait a while, and then refresh until the two modules are no longer displayed on the modules list.
    2. In your instance of Azure Percept Studio, start and view your device stream. Starting the device stream triggers a new download of HostIPModule and ImageCapturingModule that includes the latest versions.

December (2112) Release

  • Removed lines in the image frames using automatic image capture in Azure Percept Studio. This issue was introduced in the 2108 module release.
  • Security fixes for docker services running as root in azureeyemodule (mcr.microsoft.com/azureedgedevices/azureeyemodule:2112-1), azureearspeechclientmodule, and webstreammodule.

August (2108) Release

  • Azureyemodule (mcr.microsoft.com/azureedgedevices/azureeyemodule:2108-1)
    • Updated to Intel latest (May) drop for MyriadX Camera firmware update.
    • Enabled UVC (USB Video Class) camera as input source. Refer to the Advanced Development github on how to use UVC camera as input source.
    • Fixed module crash when using H.264 raw RTSP stream.

June (2106) Release

  • Azureyemodule
    • This release adds support for time-aligning the inferences of slow neural networks with the video streams. It will add latency into the video stream equal to approximately the latency of the neural network but will result in the inferences (bounding boxes for example) being drawn over the video in the appropriate locations.
    • To enable this feature, add TimeAlignRTSP: true to your module twin in the IoT Hub Azure portal.
  • Azureearspeechclientmodule
    • Integrated the Speech 1.16 SDK in the Speech module, which includes fixes for speech token support and integrates EAR SOM as default supported device in low-level lib.
    • Fixed a PnP detecting issue while security MCU removed but codec not.
    • Addressed Speech service timeouts with PTT/PTS button functions.
    • Security fixes
      • Out of Bounds Read receiving TLS data in speech module.
      • MCU and Codec USB ports are re-exposed while doing the DFU.
      • Exceptions handling when parsing JSON.
      • Enabling all hardened compiler security flags.
      • Out of Bounds Read receiving TLS data in speech module.
      • Versioning SSL and libcurl dependencies and prevent vulnerable version.
      • Enforcement of HTTPS and Pin a TLS CA on curl connections.

April (2104) Release

  • Azureyemodule
    • Fixed IoT Hub message format to be UTF-8 encoded JSON (previously it was 64-bit encoded format).
    • Fixed bug with Custom Vision classifier (previously, the Custom Vision classifier models were not working properly - they were interpreting the wrong dimension of the output tensor as the class confidences, which led to always predicting a single class, regardless of confidence).
    • Updated H.264 to use TCP instead of UDP, for Azure video analyzer integration.

Next steps