Deploy a TfLite Mobilenet quant model on Arm NN


Armnn-tflite-mobilenet-quant example uses a tensorflowlite model on top of ARMNN for image classification. In this example, the input image is a dog of malinois.
Please follow below procedure to execute this example.

Step1. Setup ARMNN environment

For program compile and execute, we should setup ARMNN environment at first. It will take long time to install packages with step by step procedure. Fortunately, we had setup ready in SP7021 development board and compatible to Respberry pi2.
If you used SP7021 board, you can directly download packages from github.


$ git clone https://github.com/sunplus-plus1/armnn_test.git


There are 4 compressed files under armnn_test/armnn-examples/armnn-pi. Please extract them and move "build" folder into "armnn" folder, finally the folder will show like below:

Please put "armnn-pi" folder in the root directory, the example programs will assume all armnn libraries put in this path.

Step2. Armnn-tflite-mobilenet-quant program download

Same to step1, the git clone armnn_test.git will also download armnn-tflite-mobilenet-quant source code. Please find the source code at armnn_test/armnn-examples/ armnn-tflite-mobilenet-quant.

Step3. Compile and execute the model

Come in the target folder
$ cd armnn-tflite-mobilenet-quant
Clean execute file
$ make clean
Compile
$ make all
Execute program
$ make test

Step4. Execute result