How to setup prepostprocessor if input tensor more than 4 dimensions in c++?

[AlloyHsu]

Hello,

I have a model with input dimensions [1,1,224,224,3].
represents
[batch, frames, height, weight, channel]

How do I set the layout of my PrePostProcessor? Is the following correct?

// Try fill cv::Mat (BGR format to input tensor

ov::preprocess::PrePostProcessor ppp(model);

ppp.input("image").tensor().set_element_type(ov::element::u8).set_color_format(ov::preprocess::ColorFormat::BGR).set_layout("N?HWC");

ppp.input().preprocess().convert_element_type(ov::element::f32);

ppp.input().preprocess().scale({ 255.0, 255.0, 255.0 });

ppp.input().preprocess().convert_layout("NCHW");

ppp.input().preprocess().convert_color(ov::preprocess::ColorFormat::RGB);

model = ppp.build();

[Vipin_S_Intel]

Hi Alloy, could you please provide us with the following details?

• The exact name and build version of the Intel® Toolkit you’re using.
• The operating system and its build version.
• Whether the product has been installed.
• A detailed explanation of your query, along with a screenshot if possible.

To assist you further, we would require these details.

[AlloyHsu]

• The exact name and build version of the Intel® Toolkit you’re using.
  w_openvino_toolkit_windows_2024.2.0.15519.5c0f38f83f6_x86_64
• The operating system and its build version.
  WIN10 22H2
• Whether the product has been installed.
  Yes
• A detailed explanation of your query, along with a screenshot if possible.
  As above

[Witold_Intel]

Hi AlloyHsu,

we have a comment from the OpenVino developers. They ask "Why do you need "frames" dimension? can it be merged with "batch"?" Could you elaborate on that please? Thanks in advance.

[AlloyHsu]

I just want porting MoViNet Video Action Recognition which create by google.

MoViNet input dimensions are as follows: [batch_size, number_of_frames, height, width, channels]

[Vipin_S_Intel]

Hi Alloy, thank you for sharing the required information. We would like to inform you that we are routing your query to the dedicated team for further assistance.

[Witold_Intel]

@AloyHsu thank you for explaining the dimensions. I will share this information with the developers.

[Witold_Intel]

Hi @AloyHsu, there is a suggestion from the devs:

> ppp.input().preprocess().convert_layout("NCHW");

 Maybe you can try ppp.input().preprocess().convert_layout("N?CHW") ? OpenVINO does not know how to convert N?HWC to NCHW, because "?" dimension is unknown. Would this be a valid solution for you?

[AlloyHsu]

ppp.input().preprocess().convert_layout("N?HWC") work for me

[Witold_Intel]

Hi Alloy, this is great news. Do you need further support from us or can we close this case?

[Witold_Intel]

Hi Alloy, do you need further support from us? I will have to close this case if there is no response for 7 business days.

[Zulkifli_Intel]

Thank you for your question. If you need any additional information from Intel, please submit a new question as this thread is no longer being monitored.