• State Resolved
  • Locked Locked
  • Replies 1 reply
  • Subscribers 100 subscribers
  • Views 1595 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

[FAQ] AM62A7: How to Use the ISP on AM62A

Jianzhong Xu
Mastermind 49605 points
Part Number: AM62A7

I want to develop camera applications on AM62A. My camera is a raw camera. How can I use the ISP on AM62A for my application? 

  • +1
    TI__Mastermind 49605 points

    Hardware IP

    The AM62Ax has an Image Signal Processor (ISP), also known as the Vision Pre-Processing Accelerator (VPAC). The specification of this hardware IP can be found in the AM62A Technical Reference Manual (TRM), chapter 7.5, Vision Pre-processing Accelerator (VPAC).

    User Interface and SW Architecture

    The AM62A ISP can be accessed by the user through either GStreamer framework or TIOVX framework. The figure below shows a simple block diagram of the image processing SW architecture on AM62A. 

    The components in the image processing pipeline are running on different cores and in different OS:

    • A53 (Linux): camera capture, 2A, applications (such as display, encoding, deep learning, etc)
    • R5 (RTOS): ISP driver

    The AM62A Linux SDK provides GStreamer plugins to access the ISP which has three HW blocks as documented in the TRM: VISS, LDC, and MSC. The Gstreamer plugins for these three blocks are:

    For all GStreamer plugins available in the AM62A SDK, please refer to https://github.com/TexasInstruments/edgeai-gst-plugins. For examples of using GStreamer to construct image processing pipeline, please refer to Edge AI dataflows.

    The AM62A SDK also provides TIOVX modules to access the ISP. For examples of using TIOVX to construct image processing pipeline, please refer to Edge AI TIOVX Apps. 

    ISP Tuning

    To effectively use the ISP on AM62A for a specific raw sensor, the ISP needs to be tuned and configured for that sensor. Please refer to the following publications for more details:

    Auto-Exposure

    The AM62A Linux SDK also provides an auto-exposure algorithm which automatically adjusts the sensor's exposure time and analog gain. This algorithm needs to know the sensor's specification on exposure time and analog gain. Please refer to this FAQ on how to pass that specification to the auto-exposure algorithm and provides the auto-exposure algorithm output back to the sensor.

    Image Flipping/Rotation

    The LDC block can be used to rotate or flip an image. Please refer to this FAQ on how to configure the LDC. 

    ISP Firmware

    The ISP/VPAC firmware can be modified and rebuilt if necessary. This firmware is not publicly available and can be requested for access from https://www.ti.com/tool/PROCESSOR-SDK-AM62A#downloads. Choose "FIRMWARE-BUILDER-AM62A" and click "Request".

    Rebuild Auto-exposure Algorithm

    TI's auto-exposure algorithm is a reference code. To rebuild or replace it, please follow this FAQ.

    Examples

    For examples of using the ISP for vision applications, please refer to the AM62A Edge AI Sample Applications.