ONNX Runtime → DirectML → Unity → Tutorial
Tutorial Links
- Training Tutorial: Train a YOLOX model using IceVision.
- Part 1: Create a dynamic link library (DLL) file in Visual Studio to perform object detection with a YOLOX model using ONNX Runtime and DirectML.
- Part 2: Perform object detection in a Unity project with ONNX Runtime and DirectML.
Demo Video
https://user-images.githubusercontent.com/9126128/185470552-bd485f66-5181-4767-b5c4-4e38d04d896d.mp4
Training Code
Note: Training on the free GPU tier for Google Colab takes approximately 11 minutes per epoch, while training on the free GPU tier for Kaggle Notebooks takes around 15 minutes per epoch.
Kaggle Datasets
Reference Images
| Class | Image |
| --------- | ------------------------------------------------------------ |
| call | ![call](./images/call.jpg) |
| dislike | ![dislike](./images/dislike.jpg) |
| fist | ![ fist](./images/fist.jpg) |
| four | ![four](./images/four.jpg) |
| like | ![ like](./images/like.jpg) |
| mute | ![ mute](./images/mute.jpg) |
| ok | ![ ok](./images/ok.jpg) |
| one | ![ one](./images/one.jpg) |
| palm | ![ palm](./images/palm.jpg) |
| peace | ![peace](./images/peace.jpg) |
| peace_inverted | ![peace_inverted](./images/peace_inverted.jpg) |
| rock | ![rock](./images/rock.jpg) |
| stop | ![stop](./images/stop.jpg) |
| stop_inverted | ![stop_inverted](./images/stop_inverted.jpg) |
| three | ![three](./images/three.jpg) |
| three2 | ![three2](./images/three2.jpg) |
| two_up | ![ two_up](./images/two_up.jpg) |
| two_up_inverted | ![two_up_inverted](./images/two_up_inverted.jpg) |