Could you please provide a detailed information about the autoexposure algorithm used in the OAK-1 AF? I would like to know what features are more relevant in the image to change the exposure parameters (ISO, exposure time) since I am having trouble in the identification of ArUco markers due to a low contrast that might be solved with a better understanding of the algorithm.

 

I check the documentation and I just found this information “By default, AutoExposure region is adjusted based on neural network output. If desired, the region can be set manually.” So, I assume that there is a dataset used to train the neural network, maybe the information I need is related to the objects in these images.

I really hope that you can help me, and thank you for your attention.

  • jakaskerl replied to this.
  • Hi @alexv
    The algorithm is closed-source.
    But this:

    alexv “By default, AutoExposure region is adjusted based on neural network output. If desired, the region can be set manually.”

    Refers to the object detection pipeline where the BBOX from NN is passed to the 3A algorithm to apply AE to best fit the ROI defined by the BBOX.

    Thanks,
    Jaka

    Hi @alexv
    The algorithm is closed-source.
    But this:

    alexv “By default, AutoExposure region is adjusted based on neural network output. If desired, the region can be set manually.”

    Refers to the object detection pipeline where the BBOX from NN is passed to the 3A algorithm to apply AE to best fit the ROI defined by the BBOX.

    Thanks,
    Jaka