In [12]:

    

from IPython.core.display import display, HTML
display(HTML("<style>.container { width:70% !important; }</style>"))


    

In [4]:

    

#Create references to important directories we will use over and over
import os, sys
current_dir = os.getcwd()
SCRIPTS_HOME_DIR = current_dir
DATA_HOME_DIR = current_dir+'/data'


    

In [5]:

    

#import modules
from utils import *
%matplotlib inline


    

    




Using TensorFlow backend.

In [6]:

    

%cd $DATA_HOME_DIR

#Set path to sample/ path if desired
path = DATA_HOME_DIR
train_path=path + '/train/binary/'
valid_path=path + '/valid/binary/'


    

    




/home/nathan/olin/spring2017/line-follower/line-follower/src/old_lane_follower_past_project/data

In [7]:

    

img_rows, img_cols = 16, 32
in_shape = (img_rows, img_cols)
batch_size = 64
nb_classes = 3


    

In [13]:

    

gen = image.ImageDataGenerator(rescale=1. / 255) #from keras preprocessing. 
#generates batches of tensor image data, data will loop in batches indefinitely
#rescales data before doing any other iteration (here, will rescale pixel intensities from 0-255 to 0-1)

#what does this "looping in batches indefinitely" actually mean?


    

In [9]:

    

train_batches = get_batches(train_path, batch_size=batch_size, 
                            target_size=in_shape, color_mode="grayscale", 
                            gen=gen)
val_batches   = get_batches(valid_path, batch_size=batch_size, 
                            target_size=in_shape, color_mode="grayscale", 
                            gen=gen)


    

    




Found 938 images belonging to 3 classes.
Found 536 images belonging to 3 classes.

In [ ]:

    

model = Sequential([
    Convolution2D(32,3,3, border_mode='same', activation='relu', input_shape=(img_rows, img_cols, 1)),
    MaxPooling2D(),
    Convolution2D(64,3,3, border_mode='same', activation='relu'),
    MaxPooling2D(),
    Flatten(),
    Dense(1024, activation='relu'),
    Dense(nb_classes, activation='softmax')
    ])
model.compile(Adam(), loss='categorical_crossentropy', metrics=['accuracy'])


    

In [ ]: