Error Init: Unable to find the specified file about hololens2-machine-learning HOT 7 OPEN

Thank you very much @doughtmw for the quick reply. I have also got it working. @NTUZZH I trained my custom model via ML.net through Azure ML (Image Classification) and got an ONNX file. The we can use the sample from Mr. doughtmw, and adapt the scripts according to our model. The only place now I am checking is whether the labelled data must be a Json file or can it be done also with TXT file. I just have one more question @doughtmw, whether the sample can also be adapted for Object Detection Model and could you please give some references or any github sample where by an object detection is carried out with an ONNX file locally in HL2 itself? Since I am kind of new to coding, I am not sure how to code the Bounding Boxes. Please do help me with this issue.

from hololens2-machine-learning.

SAME PROBLEM HERE. How to deal with this Error?

from hololens2-machine-learning.

Hi @Avinash0795 and @NTUZZH, the HL2 build version could be a factor but a more likely issue could be due to incorrect formatting of the ONNX model in Visual Studio. Be sure to follow all steps in the README for copying the ONNX model into the correct location after building your Visual Studio project as well as setting the ONNX model properties. Here's a similar recent issue on this repo with a suggested solution.

from hololens2-machine-learning.

Dear @doughtmw ,
Yes I reset the path of the model.onnx and make it inside of Asset file, then it works! So yes please make sure the model.onnx to be inside of Asset file. Thanks a lot!

May i ask for your suggestions, what should i do if i want to use my own dataset to recognize the images? Should I first train the images through the tensorflow and then import the training dataset as a ONNX file?

from hololens2-machine-learning.

Thank you very much @doughtmw for the quick reply. I have also got it working. @NTUZZH I trained my custom model via ML.net through Azure ML (Image Classification) and got an ONNX file. The we can use the sample from Mr. doughtmw, and adapt the scripts according to our model. The only place now I am checking is whether the labelled data must be a Json file or can it be done also with TXT file. I just have one more question @doughtmw, whether the sample can also be adapted for Object Detection Model and could you please give some references or any github sample where by an object detection is carried out with an ONNX file locally in HL2 itself? Since I am kind of new to coding, I am not sure how to code the Bounding Boxes. Please do help me with this issue.

Hi @Avinash0795, noted and thanks for your information. I'm just like you and some kind of new hand to C# coding. So i'm also very interested in your issue that how to use the Object Recognition to give the object of interest the Bounding Boxes.

And @doughtmw thank u sir.

from hololens2-machine-learning.

Respected Sir,

I have another Question. Will this work even if the Application is setup for MixedReality. Because somehow
after I included the Mixed Reality Toolkit to the Scene, the Text is not being seen. Is there any Idea regarding
this issue. I tried changing the Text to Text Meshpro but still somehow it didnot work out.

It would be great if you could guide me through this.

from hololens2-machine-learning.

Hi , I just tried running the sample , I do copy and paste the ONNX file to the builds/assets, but when setting up the properties of that file, there is no file type "document". Is there any alternative file type to use?

from hololens2-machine-learning.