VGG_new.py

from __future__ import print_function

import numpy as np
from random import shuffle
import time
import csv
from PIL import Image
import os

import tensorflow as tf
import VGG
import keras
from keras.callbacks import EarlyStopping, LearningRateScheduler
from keras import initializers
from keras.preprocessing import sequence
from keras.utils import np_utils
from keras.models import Sequential,load_model,Model
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import *
from keras.callbacks import CSVLogger
from keras import callbacks
from keras.preprocessing.image import ImageDataGenerator

from sklearn.metrics import classification_report
import sklearn.metrics as sklm

np.random.seed(3768)  # for reproducibility

#Get TensorFlow session
def get_session(): 
  config = tf.ConfigProto() 
  config.gpu_options.allow_growth = True 
  return tf.Session(config=config) 

# use this environment flag to change which GPU to use 
os.environ["CUDA_VISIBLE_DEVICES"] = "0" 

# set the modified tf session as backend in keras 
keras.backend.tensorflow_backend.set_session(get_session())

# One hot encoding of labels  
def dense_to_one_hot(labels_dense,num_clases=5):
  return np.eye(num_clases)[labels_dense]

# Loading training images
def loadTrainImageAndLabels():
    imgList=[]
    labelList=[]
    reader = open("/home/reza/DeepKneeExplainer/resources/newlabel.csv") #label file path
    
    data=reader.readlines()
    files = os.listdir('/home/reza/DeepKneeExplainer/resources/Data/MRI/unbalanced/up/training/') #training path
    
    shuffle(files)
    
    for file in files:
        if file.endswith(".xml"):continue
        fi_d = os.path.join('/home/reza/DeepKneeExplainer/resources/Data/MRI/unbalanced/up/training/',file)#training path
        img=Image.open(fi_d)
        im=np.array(img.resize((360,360), Image.ANTIALIAS))
        patient=file.split('_')[0]
        direction=file.split('_')[1].split('.')[0]
        label="q"
        for row in data:
            if patient in row.split(",")[0]:
                if "L" in direction:
                    label=row.split(",")[1]
                else:
                    label=row.split(",")[4]
                break
        if "8" not in label and "9" not in label:
            label= dense_to_one_hot(int(label),5)
            imgList.append(im)
            labelList.append(label)
    return np.array(imgList),np.array(labelList)

# Loading validation images
def loadTestImages():
    imgList=[]
    labelList=[]
    reader = open("/home/reza/DeepKneeExplainer/resources/newlabel.csv") #label file path
    
    data=reader.readlines()
    files = os.listdir('/home/reza/DeepKneeExplainer/resources/Data/MRI/unbalanced/up/validation/') #test path
    
    for file in files:
        if file.endswith(".xml"):continue
        fi_d = os.path.join('/home/reza/DeepKneeExplainer/resources/Data/MRI/unbalanced/up/validation/',file) #test path
        img=Image.open(fi_d)

        im=np.array(img.resize((360,360), Image.ANTIALIAS))
        patient=file.split('_')[0]
        direction=file.split('_')[1].split('.')[0]
        label="q"

        for row in data:
            if patient in row.split(",")[0]:
                if "L" in direction:
                    label=row.split(",")[1]
                else:
                    label=row.split(",")[4]
                break
        if "8" not in label and "9" not in label:
            label= dense_to_one_hot(int(label),5)
            imgList.append(im)
            labelList.append(label)
    return np.array(imgList),np.array(labelList)

# Loading the labels of validation images
def loadTestLabels(): #load labels with decimal format for comparison
  imgList=[]
  labelList=[]
  reader = open("/home/reza/DeepKneeExplainer/resources/newlabel.csv") #label file path
  data=reader.readlines()
  files = os.listdir('/home/reza/DeepKneeExplainer/resources/Data/MRI/unbalanced/up/validation/') #test path

  for file in files:
        if file.endswith(".xml"):continue
        patient=file.split('_')[0]
        direction=file.split('_')[1].split('.')[0]
        label="q"

        for row in data:
           if patient in row.split(",")[0]:
              if "L" in direction:
                 label=row.split(",")[1]
              else:
                 label=row.split(",")[4]
              break
        if "8" not in label and "9" not in label:
          labelList.append(int(label))
  return np.array(labelList)

#Defining hyperparameters
batch_size=32

#Instantating VGG19 model
model = VGG.VGG19_dense((360,360,1),5) #VGG19_dense for revised VGG19, VGG19 for VGG19. Please pay attention to VGG16(), chnage the input shape and class number in VGG.py.

#Creating an optimizers
adagrad = keras.optimizers.Adagrad(lr=0.01)
model.compile(optimizer=adagrad, loss='categorical_crossentropy',metrics=['accuracy'])

#Image data generation for the training 
datagen = ImageDataGenerator(
               featurewise_center=True, 
               samplewise_center=False,  # set each sample mean to 0
               featurewise_std_normalization=True,  
               samplewise_std_normalization=False)  

# Preparing training and test sets
X_train, Y_train = loadTrainImageAndLabels()
X_test, Y_test = loadTestImages()

X_train = X_train.reshape( len(X_train), len(X_train[0]), len(X_train[0][0]),1)
X_test = X_test.reshape( len(X_test), len(X_test[0]), len(X_test[0][0]),1)

Y_train = Y_train.reshape(len(Y_train), 1, 1,5)
Y_test = Y_test.reshape(len(Y_test), 1, 1,5)
X_train = X_train.astype('float32')
X_test = X_test.astype('float32')

X_train /= 255
X_test /= 255

datagen.fit(X_train) 
for i in range(len(X_test)):
      X_test[i] = datagen.standardize(X_test[i])

#Creating early stopping 
earlystop=EarlyStopping(monitor='val_acc', min_delta=0, patience=200, verbose=1, mode='auto', restore_best_weights=True)       

# Model training 
numEpochs=500
history = model.fit_generator(datagen.flow(X_train, Y_train,batch_size=batch_size),steps_per_epoch=32,epochs=numEpochs,shuffle=True,validation_data=(X_test, Y_test), verbose=1)

from lossprettifier import LossPrettifier
reporter = LossPrettifier(show_percentage=True)

# visualizing losses and accuracy
train_loss = history.history['loss']
val_loss = history.history['val_loss']
#print(train_loss)
#print(val_loss)

for i in range(numEpochs-1):
# Some Training Functions 
    reporter(epoch=i, LossA = train_loss[i], LossAB = val_loss[i])

# Model evaluation 
score, acc = model.evaluate(X_test,Y_test,batch_size=batch_size)
print("Accuracy:",acc)

#if acc>0.675:
model.save_weights("VGG19-KLnew.h5")

y_pred = model.predict(X_test)
y_pred = y_pred.reshape( len(Y_test), 5)
Y_predict = y_pred.argmax(axis=-1)

# Writing results on file
f=open('VGGRESULTS-KLnew_updated.txt','a') #create classification report
f.write(classification_report(loadTestLabels(), Y_predict))
f.write(str(sklm.cohen_kappa_score(loadTestLabels(), Y_predict))+","+str(acc)+","+str(score)+"\n")

#Print class-wise classification metrics
print(classification_report(loadTestLabels(), Y_predict))