wrong dimensions from yolov3

[hoaquocphan]

Hello All,

currently I import the yolov3.onnx model to onnx v1.1.0 but it have below issues:

there are 2 input, and the dimensions 2 and 3 of 2 inputs is always -1
input 1:
dim 2=-1
dim 3=-1
input 2:
dim 2=-1
dim 3=-1

when importing yolov2.onnx model to onnx v1.1.0, the dimensions is always as below:
dim 2=416
dim 3=416

I don't know what is "-1" mean, and how to show correct dimensions.
so Please help me to detect the issue with my problem

Thanks

[hoaquocphan]

Hi Mrs @sowson
I'm a newbie in AI, so could you share me a example .cpp code to import yolov3 model on onnxruntime.

I have searched on GG but there are no any thing to apply
My project have deadline at end this week, so please help me.

Thanks

[sowson]

hmm, I think I can help you, pls go to one of the closed issue: #12 and read it carefully, if you have more questions after that I will be happy to help you ;-). But I think this task will make you a little bit more convenient with your project! Thanks!

[hoaquocphan]

Hi @sowson
Thanks for your reply.

I would like to give you some more detail about this:

• I don't work with Yolo v2, I already created an .cpp file to port to onnxruntime
• I port the model TinyYolo v3 from onnx model https://github.com/onnx/models/tree/master/vision/object_detection_segmentation/tiny_yolov3
  not train it.

There are some sample Preprocessing,Postprocessing steps in python code to port it, but I haven't work with Python so I'm creating an .cpp file to port it and have some difficulty.

So if you have some sample Preprocessing,Postprocessing steps in .cpp to port tiny Yolo v3
Please share it for me.

Thanks

[sowson]

Hi, @hoaquocphan I am sorry I do not know how to help you with these tasks, anyone?

[hoaquocphan]

Hi Mr @sowson

Thank you very much for your help.
I found the rootcause of wrong dimension, it because define in the source code of keras-yolo3