Implement Trained Weight in Android

[amar-mustaqim]

Hi. I'm trying to develop an android camera application that able to detect sign language. But, how to implement my own dataset that i've trained with the android application that i want to build?

[dreambit]

Hi. I would get started building backend(there is darknet.py that u can turn into http server) that accepts images over http and returns result to your android app.

[AlexeyAB]

@amar-mustaqim Hi,

If you can use OpenCV for Android 3.4.0 or higher https://opencv.org/releases.html in your application (for example https://sourceforge.net/projects/opencvlibrary/files/4.0.1/opencv-4.0.1-android-sdk.zip/download ) then you can just use these examples of Yolo v3, especially these lines from examples:

• C++: https://github.com/opencv/opencv/blob/55f02d8c66baf0747805f75762bdaca2e2d1985e/samples/dnn/object_detection.cpp#L192-L221

• Python: https://github.com/opencv/opencv/blob/55f02d8c66baf0747805f75762bdaca2e2d1985e/samples/dnn/object_detection.py#L129-L150

---

Or you can use this repository that is optimized for ARM CPUs: https://github.com/shizukachan/darknet-nnpack It supports YOLOv3-Tiny, but I don't know whether it supports YOLOv3-Full model.

[amar-mustaqim]

Thank you so much! Will figure it out

[spaul13]

@amar-mustaqim @dreambit do u guys figure it out how to directly run the trained darknet model on android? If so, it will great help if u can instruct me.

[AlexeyAB]

@spaul13

You can try to use Yolov3 with TVM framework on Android: https://docs.tvm.ai/tutorials/frontend/deploy_model_on_android.html

Or look at other frameworks: https://github.com/AlexeyAB/darknet#yolo-v3-in-other-frameworks

[spaul13]

Thanks a lot @AlexyAB. Can u plz help me in more thing can u plz tell me how can I implement reinforcement learning for classification using darknet?