This tutorial will show how to implement Deep Neural Network for pixel based supervised classification of Sentinel-2 multispectral images using keras package in R under Windows 10.
keras is a popular Python package for deep neural networks with multiple backends, including TensorFlow, Microsoft Cognitive Toolkit (CNTK), and Theano. Two R packages allow you to use [Keras[(https://keras.rstudio.com/)] from R: keras and kerasR. The keras package is able to provide a flexible and feature-rich API and can run both CPU and GUP version of TensorFlow in both Windows and Linux. If you want to run this tutorial with GUP version of TensorFlow you need following prerequisites in your system:
*NVIDIA GUP: First, you must make sure weather your computer is running with NVIDIA® GPU or not. Follow the instruction as described here.
*CUDA Toolkit v9.0: If you have an NVIDIA® GPU in your system, you need to download and install CUDA Toolkit v9.0. Detail installation steps can be found here.
*cuDNN v7.0: The download the zip file version cuDNN v7.0 for your CUDA Toolkit v9.0.You need to extract the zip file and add the location where you extracted it to your system PATH. Detail installation steps can be found here [here](F:\DeepLearning_tutorial\Satellite_Image_Calssification\h20_R_ImageCalssification\keras_R\Detail installation steps are described here).
Detail installation steps of Keras backend GPU or CUP version of Tensorflow can be found here.
First, we will split "point_data" into a training set (75% of the data), a validation set (12%) and a test set (13%) data.The validation data set will be used to optimize the model parameters during training process.The model's performance will be tested with the test data set and then we will predict landuse clasess on grid data set. The point and grid data can be download as rar, 7z and zip format.