How to do object detection in C# using yolov5?

See more:

C#

ML.Net

I am new to machine learning and stuff.
I wanted to do custom object detection in C# using webcam at runtime.

The thing is as per client request I did this through roboflow but client wanted fully offline since roboflow did not provide this feature, I looked into yolo. Although roboflow fulfilled object detection but offline feature meant I had to find some other way to do this. To try new method I downloaded small public dataset for balloons from roboflow in yolov5pytorch format. this dataset has blue and red balloons as classes.

Now I just needed to train this and implement it in C# Winform project.

I opened google collab yolov5 to train my dataset, I followed each step and after training I checked it on video and it was detecting red and blue balloons.

Now my question was how to implement this custom trained model in my C# code for object detection.

Since I have yolov5 custom trained model which tutorial to follow to implement this.

I tried to download yolov5 in onnx format and tried it with ml.net but did not understand, the guy was working with custom vision onnx model so followed each step which failed. I do not understand all depth of how it is working.

So now I am stuck with trained model in yolov5 in google colab. What steps should I follow or any tutorial link to integrate this in my c# project?

Also can I use this trained yolov5 model in python and intrgrate python script in c#. Remember it has to detect at runtime and offline.

What I have tried:

Tried roboflow (no offline feature)
Trained custom dataset on google colab and downloaded onnx model of yolov5 and implement it in winform project with help of ml.net.