[SOLVED] Snapdragon Neural Processing Engine (SNPE) and Tiny Yolo on Android App

Snapdragon Neural Processing Engine (SNPE) and Tiny Yolo on Android App

for my final university exam I am trying to use SNPE with Tiny YOLO for real time object detection in an Android App. I succesfully converted model to DLC format, but i can't understand how to prepare input tensors and how to process output tensors. Can sameone help me? Thanks.

Solution

Steps to build SNPE neural network and get the output FloatTensor:

Create an asset folder in Android/app directory and keep the model file(.dlc) in the asset folder.

// assetFileName is the file name of .dlc
InputStream assetInputStream = application.getAssets().open(assetFileName); // Create and build the neural network
NeuralNetwork network = new SNPE.NeuralNetworkBuilder(application)
        .setDebugEnabled(false)
//outputLayerNames can be got while converted model to DLC format
        .setOutputLayers(outputLayerNames)
        .setModel(assetInputStream, assetInputStream.available())
        .setPerformanceProfile(NeuralNetwork.PerformanceProfile.DEFAULT)
        .setRuntimeOrder(selectedRuntime) // Runtime.DSP, Runtime.GPU_FLOAT16, Runtime.GPU, Runtime.CPU
        .setCpuFallbackEnabled(needsCpuFallback)
        .build();
// Close input
assetInputStream.close();

Create an Input Tensor
Propagate Input Tensors Through the Network
Process the Neural Network Output

Please follow the link below and find sections mentioned in steps 2,3 and 4 for preparing input Tensors and processing output tensors https://developer.qualcomm.com/docs/snpe/android_tutorial.html