GCN_process.py

def applyGCN(save_image=False,doTrain=False,doTest=False):
    data = np.loadtxt("../inputData/train.csv", dtype=np.float32, delimiter=',', skiprows=1)
    test_data = np.loadtxt("../inputData/test.csv", dtype=np.float32, delimiter=',', skiprows=1)
 
    trainingFolder = "../inputData/converted_training/GCN/"
    testingFolder = "../inputData/converted_testing/GCN/"
 
    if(doTrain):
        aux_data = data.copy()
 
        #############TRAIN_DATA############################
 
        data_to_gcn = data[:, 1:]
        labels = data[:, 0]
 
        img_gcn = global_contrast_normalize(data_to_gcn)
 
        print "Saving training images"
        with open(trainingFolder + "GCN_traindata.csv", 'wb') as fp:
           for x in xrange(0, len(data[:, 1])):
           #for x in xrange(0, 1):
               image_aux = img_gcn[x, :].copy()
               print "size ",image_aux.size
               print "np.max ",np.max(image_aux)
               print "np.min ",np.min(image_aux)
 
               #aux_data[x, 1:] = image_aux.tolist()
               image = np.reshape(image_aux, (28,28))
 
               if save_image is True:
                   mpimg.imsave(trainingFolder + "GCN_Gray_TrainImage_" + str(x), image, cmap=plt.get_cmap('gray'))
               line = image_aux.tolist()
               line = reduce(lambda x,y:str(x)+','+str(y), line)+'\n'
               line = str(labels[x])+line
               fp.write(line)