• State
  • Locked Locked
  • Replies 11 replies
  • Subscribers 99 subscribers
  • Views 472 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-J784S4: Scaler node replication using multiple graphs LDCs outputs as input.

Lalit Meena
Intellectual 410 points
Part Number: PROCESSOR-SDK-J784S4


Hi TI,

Need help regarding scaler node replication.

Scenario: I have 3 graphs.

G1 : capture1->viss1->ldc1->scaler1->........

G2: capture2->viss2->ldc2->scaler2->........

G3: scaler3-->node_x->node_y->.....

G3's first node is scaler3 node.

Goal : I want to give ldc1 and ldc2 output to the scaler3 node i.e. two input images to the scaler3 node. Both ldc1 and ldc2 outputs are objects arrays. How can I create an object array having both of these outputs to give in to scaler3.

thanks and regards,

Lalit

  • 0
    TI__Guru 52763 points

    Hi Lalit,

    Could we get a clarification of whether the graphs are created using OpenVX nodes or GStreamer nodes? Additionally, I assume this is Processor SDK J784S4 Linux+RTOS but could you confirm?

    Regards,

    Takuma

  • 0 in reply to Takuma Fujiwara
    Intellectual 410 points

    Hi Takuma Fujiwara,

    I am using OpenVX framework for all the graphs and sdk is QNX+RTOS not linux. 

    Thanks and regards.

  • 0 in reply to Lalit Meena
    TI__Guru 52763 points

    Hi Lalit,

    Understood. Let me forward this to a colleague of mine to see if he can comment on the graphs. Please expect a response within 1~2 business days.

    Regards,

    Takuma

  • 0 in reply to Takuma Fujiwara
    Intellectual 410 points

    Thank you. Please reply as soon you can as this is a critical task at my end and I'm stuck on this.

  • 0 in reply to Lalit Meena
    TI__Guru* 85666 points

    Hi,

    In the graph3, what is the output of the scalar node? Will it scale down and combine the 2 outputs of ldc nodes and give a single output to node_x?

    Regards,

    Nikhil

  • 0 in reply to Nikhil Dasan
    Intellectual 410 points

    Hi Nikhil,

    Graph3 is an OD graph. It has pre proc node, tidl node, draw box detection node. So the first output of scaler3 is used by the tidl node and second output is used by draw box detection node.

    regards,

    Lalit

  • 0 in reply to Lalit Meena
    TI__Guru* 85666 points

    Hi Lalit,

    Then what is the significance of 2 LDC outputs and combining them to one object array to input to Scalar node?

    If the scalar is being fed the 0th index of the object array and then replicate node happens, then why do we have 2 ldc node outputs?

    Regards,

    Nikhil

  • 0 in reply to Nikhil Dasan
    Intellectual 410 points

    Hi Nikhil,

    We are doing batch processing in OD graph. Two different camera images(Driver Side image G1 graph, Passenger side image, G2 graph) are need to be processed in OD graph G3.

    regards,

    Lalit

  • 0 in reply to Lalit Meena
    TI__Guru* 85666 points

    Hi Lalit,

    The issue here is that an object array should contain identical resolution and datatype of images.

    Since you have 2 different camera images (I am assuming different resolution as you have used 2 capture nodes), this cannot be combined into a single object array.

    Since the multiscalar node takes in only 1 input, you would have to use 2 MSC nodes here.

    Or, if the data-format is same, the you can use only 1 capture node pipeline and send the ldc output to graph 3 scalar node by making both as graph parameters.

    Regards,

    Nikhil

  • 0 in reply to Nikhil Dasan
    Intellectual 410 points

    Hi Nikhil, 

    Can we achieve this if both the image's resolutions are same. Or we have to use only one capture node pipeline to achieve this in that case also. We are using two different capture nodes(two different CSI-RX instances) because of functional safety requirements.

    So my question is, can we not create an object array(for input to scaler3 of graph G3) using two object arrays(LDC1 and LDC2 out arr)? The two object arrays are containing similar type of objects(vx_image in my case) and have same resolution.

    regards,
    Lalit

  • 0 in reply to Lalit Meena
    TI__Guru* 85666 points

    Hi Lalit,

    This should be possible, but for this you would have to create a node or a logic in the application, where you could combine 2 object arrays to one with similar resolution. 

    This is not available currently in the SDK, but you take the reference of the object array splitter node and implement an object array combiner node.

    You could refer the implementation of the object array splitter node in the file tiovx/kernels/openvx-ext/target/vx_obj_array_split_target.c in SDK 9.1

    Regards,

    Nikhil