search

TakahiroMiyaura / HoloLensOnnxTinyYoLo

This project is a sample code that evaluate iamge data using "Tiny Yolo V2" on Window ML(RS4/RS5)
MIT License
4 stars 1 forks source link

having problems with namespaces #1

Open valavakilian opened 5 years ago

valavakilian commented 5 years ago

Hi and thanks for sharing your work. I was trying to run this application on the Hololens but I run into the following errors (I couldn't run the application on unity runtime version .NET 3.5 so I changed it to 4.6) after I try to build it on unity :

error CS0246: The type or namespace name `Windows' could not be found. Are you missing an assembly reference?

error CS0246: The type or namespace name `TensorFloat' could not be found. Are you missing an assembly reference?

error CS0246: The type or namespace name `LearningModel' could not be found. Are you missing an assembly reference?

error CS0246: The type or namespace name `IRandomAccessStreamReference' could not be found. Are you missing an assembly reference?

I checked the visual studio file and the following libraries are not imported:

using Windows.Media; using Windows.Storage; using Windows.Storage.Streams; using Windows.AI.MachineLearning;

I have run into this problem previously as well and I couldn't solve it. I'm not sure but I assume there is something wrong with my SDK as I don't have Windows Insider preview.

Thank you for your help

TakahiroMiyaura commented 5 years ago

Hi valavakilian.

Thank you for letting me know. Cloning this Repository, I confirmed a similar phenomenon. When I tested on local environment at the time,I could build these UWP Projects :_(

So, I update projects and confirmed build success. I changed the structure of the project from last time.

If you don't mind, could you please try it one more time?

valavakilian commented 5 years ago

Hi and thanks for the reply. I am currently trying to use this updated version but I still run into the same errors. I can simply import :

using Windows.Media; using Windows.Storage; using Windows.Storage.Streams; using Windows.AI.MachineLearning;

these libraries in a blank UWP application but for some reason I can't import them when using your unity project. I simply go to your 1803 and 1809 folder after downloading the whole repo and try to build the unity program when I run into errors.

thanks for your help

TakahiroMiyaura commented 5 years ago

Hi valavakilian.

Could it be that you build these programs with Unity? if it is yes,I didn't understand it.

This samples made with Visual Studio 2017. This application is 2D UWP, so we don't need build with Unity.

Could you please open and build the solution using Visual Studio 2017. So, you will be able to deploy and execute on HoloLens!

thanks,

valavakilian commented 5 years ago

Hi and thanks for your help . Yes indeed I was trying to make the application with Unity. Sorry that was totally my bad. I though I would be able to use the scripts in unity. I was able to successfully create the project on hololens but there are some out-of-bounds problems when I run the application on the hololens itself. I will try to take a look at them and see if there is anything wrong with it but considering that you were able to build it successfully it is possibly something from my side. Also, I am interested to know if you think it is possible to implement the same structure for live video taken from the camera stream in real time. To use tiny yolo to give a description of detected objects in the hololens field of view with a low frame rate ? I have searched for a while to see if it is possible to include yolo in hololens with unity but it seems like winRT and unity are not compatible at all.

Thanks for your help and amazing work

TakahiroMiyaura commented 5 years ago

Hi valavakilian.

I’m grad that your problem has been solved!

Also, I am interested to know if you think it is possible to implement the same structure for live video taken from the camera stream in real time.

Well,I I’ve implemented “Object Detection Hololens App” using ONNX in real time, There ware low performance than I expected. I’ve tried bellow aproach,

To use tiny yolo to give a description of detected objects in the hololens field of view with a low frame rate ?

There is about over 2-3sec to evaluate TinyYoloV2 in HoloLens:_( For this reason , I think that it is difficult to evaluate using Onnx in real time…

I have searched for a while to see if it is possible to include yolo in hololens with unity but it seems like winRT and unity are not compatible at all.

I've implemented 3D application that use any ONNX models with Unity and Visual Studio2017.

Also, I've modified "MR and Azure 310: Object detection" to use ONNX that export from Custom vision Service.

This sample codes is bellow: https://github.com/TakahiroMiyaura/ObjDetectionSamplesUsingONNX

blog(Sorryr, Japanese Only...)

Thanks!