• State
  • Locked Locked
  • Replies 5 replies
  • Subscribers 98 subscribers
  • Views 362 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

PROCESSOR-SDK-AM69A: We have a custom script for image analysis running on an intel core i7 processor. Would an AM69A perform faster and better for image analysis?

Stefan Vermeulen
Prodigy 20 points
Part Number: PROCESSOR-SDK-AM69A
Other Parts Discussed in Thread: AM69A

Tool/software:

We currently do image analysis with a 50  mp camera. This is done in a timelapse fashion. With our current hard and software (windows, pythonscript) it takes around 5-10 minutes. We have to recognise and track around 20 to 40 000 features which are circles in particular segments of the images in function of time. The current algorithm we use is based on Hough transform. We want to improve on our time of analysis, increase our camera to a 250 mp camera and increase the tracking performance. Would the EdgeAI feature be a good choice to implement.

  • 0
    TI__Guru* 86786 points

    Hi,

    The AM69A is a 32 TOPS device which has 3 CSI2-RX capture (which supports upto 10Gbps capture rate over the MIPI lines) and have 2 Video processing accelarators to process the same.

    AM69x Processors, Silicon Revision 1.0 datasheet (Rev. C) (ti.com)

    Based on the above, I would to get a bit more brief about your usecase here.

    i.e.

    What is capture format here?,

    Is the 250 MP camera outputs a csirx frame format? 

    What is the frame rate at which the data is captured?

    Could you elaborate more on your usecase?

    Regards,

    Nikhil

  • 0 in reply to Nikhil Dasan
    Prodigy 20 points

    Hello Nikhil, 

    The camera output is distinct images.

    current framerates are one frame each 5 minutes.

    The current postprocessing untill result is 5 minutes so we are limited currently by this processing time

    current sensor resolution: 45.7 megapixel
    Maximal fotoresolution: 8256 x 5504 pixels
    image formats: JPEG, TIFF, NEF(RAW), DCF 2.0, Exif 2.31, PictBridge

    Sensor resolution: using a 61 megapixel camera with a 16 frame pixel shift we get 240.8 mega pixels
    Maximal fotoresolution: 19008 x 12672 pixels
    image formats: JPEG (DCF Ver. 2.0, Exif Ver.2.31, MPF Baseline compliant), RAW (Sony ARW 2.3 format)

    I checked and did not find csirx as a format in the specs. We do not have a video feed.

    240.8 mp would increase our process times by 5 fold

    regards

    Stefan 

  • 0 in reply to Stefan Vermeulen
    TI__Guru* 86786 points

    Hi Stefan,

    Thank you for the details,

     So you are not using a sensor that outputs frames using csi2 protocol.

    In this case what is the mode of connection or transfer protocol you are looking here?

    Is it USB or Ethernet or something else?

    Regards,

    Nikhil

  • 0 in reply to Nikhil Dasan
    Prodigy 20 points

    Hello Nikhil,

    Sorry for the incompleteness of my answers. We use USB but ethernet can also be used?

    regards Stefan

  • 0 in reply to Stefan Vermeulen
    TI__Guru* 86786 points

    Hi Stefan,

    Let me check internally with the respective to USB and ethernet inputs. We do support both these inputs on the board.

    Stefan Vermeulen said:
    The current algorithm we use is based on Hough transform

    I believe you would be using our DSPs for this? 

    Regards,

    Nikhil