train.py

# -*- coding: utf-8 -*-
"""
Created on Mon Feb  5 10:36:33 2018

@author: IbrahimD
"""

from keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D, Conv2DTranspose, concatenate
from keras.models import Model, Sequential
from keras import backend as K
from keras.optimizers import Adam
from keras.callbacks import ModelCheckpoint, EarlyStopping

import glob
import sys
import matplotlib.pyplot as plt
import numpy as np
import pickle

import unet, Inception, unetV2
import read_data

from keras.models import load_model  
from keras.preprocessing.image import ImageDataGenerator


def save_history(hist, filename):
    
    with open(filename, 'wb') as file_pi:
        pickle.dump(hist.history, file_pi)

def dice_coef(y_true, y_pred):
    y_true_f = K.flatten(y_true)
    y_pred_f = K.flatten(y_pred)
    intersection = K.sum(y_true_f * y_pred_f)
    return (2. * intersection + smooth) / (K.sum(y_true_f) + K.sum(y_pred_f) + smooth)


def dice_coef_loss(y_true, y_pred):
    return -dice_coef(y_true, y_pred)

smooth = 1.


MASSACHUSETTS_PATH = "Massachusetts/"
TRAINING_SET = 1
MODEL_NAME = 'UNETV2' # or 'UNET' or 'INCEPTION'

window = 28 * 8


path = MASSACHUSETTS_PATH + 'train/'
x_train, y_train = read_data.read(path, 110)

if 2 == TRAINING_SET:
    index = 75 * 49
    x_train = x_train[0:index,:,:,:]
    y_train = y_train[0:index,:,:,:]


print("len train ", len(x_train))

path = MASSACHUSETTS_PATH + 'validation/'
x_valid, y_valid = read_data.read(path, 4)
print("len valid ", len(x_valid))


if 'UNET' == MODEL_NAME:
    model = unet.get_unet()
if 'INCEPTION' == MODEL_NAME:
    model = Inception.get_unet()
if 'UNETV2' == MODEL_NAME:
    model = unetV2.get_unet_plus_inception()


model.compile(optimizer=Adam(lr=1e-5), loss=dice_coef_loss, metrics=[dice_coef])


model_name = MODEL_NAME
save_weights_path = "results-1/" + model_name


orig_stdout = sys.stdout
f = open('ModelSummary.txt', 'w')
sys.stdout = f
print(model.summary())
sys.stdout = orig_stdout
f.close()


epoch = 100


strTemp = save_weights_path + model_name + ".h5"


mc = ModelCheckpoint(strTemp.replace('.h5','.weights'), mode='min', save_best_only=True, save_weights_only=True)
es = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=10)
callbacks_list = [mc, es]


history = model.fit([x_train], [y_train],
                               validation_data=(x_valid, y_valid), 
                               callbacks=callbacks_list,
                               batch_size=5,
                               epochs=epoch)



strTemp = save_weights_path + model_name + ".history"
save_history(history,strTemp)