Error Init: Unable to find the specified file

[Avinash0795]

Respected Sir, I have followed the each and every step mentioned in the Github Sample and after the Deployment into HL2, the application is not detecting the sample Image provided. Could this be an issue due to the HL2 Build Version. I am using a Higher Build 20348, Windows Holographic version 21H2. I understand that the application is not getting access to the HL2 Camera too. Is there any constraints to the size of the Image being displayed on the Screen? Please do help me with this issue. Sincerely, Avinash Shanmugam

[NTUZZH]

SAME PROBLEM HERE. How to deal with this Error?

[doughtmw]

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.

[NTUZZH]

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?

[Avinash0795]

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.

[NTUZZH]

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.

[Avinash0795]

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.

[pshiri]

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?