Building is done by cmake. To successfully build the project you need to have c++14 capable compiler and opencv-2.4 library installed.

Following steps could be useful for usage of neural network:

In order to download data set use download.sh script from directory where you want the data to be stored. download.sh takes two arguments of time interval of samples. Example:

    ./download.sh 2015-1-1 2015-12-31

would download the whole year 2015.

If you want to change the sample rate, rewrite the line 22 in download script to desired rate in minutes.

pca-matrix binary is used to compute PCA matrix from dataset. Usage example:

    ./pca-matrix my-dataset my-dataset.pca

Where my-dataset is directory located in same directory as 'pca-matrix' binary and contains only downloaded data.

To run lstm network in learning mode run following command:

    ./lstm --data=[data set path] --pca=[path to pca matrix] --ite=[number of iterations]
           --s=[save network path]

Before run be sure, that the dimension of pca matrix is same as input_size in main.cpp.

If you dont know the pca matrix size, run the lstm binary with given matrix and change the input_size on line 22 in main.cpp.

To run lstm network in evaluation mode run following command:

    ./lstm --data=[data set path] --pca=[path to pca matrix] --l=[network path] --o=[result.jpg]

The network will load first 7 images from data set and computes the following one.