├── README.md
├── predict_image.py
└── Lungs.py


/README.md:
--------------------------------------------------------------------------------
1 | # Deep-Learning-in-Medical-Science


--------------------------------------------------------------------------------
/predict_image.py:
--------------------------------------------------------------------------------
 1 | ## Predicitng Models
 2 | 
 3 | 
 4 | from keras.models import load_model
 5 | from keras.preprocessing import image
 6 | from keras.applications.vgg16 import preprocess_input
 7 | import numpy as np
 8 | model = load_model('model_vgg19.h5')
 9 | img = image.load_img('val/PNEUMONIA/person1946_bacteria_4874.jpeg', target_size=(224, 224))
10 | x = image.img_to_array(img)
11 | x = np.expand_dims(x, axis=0)
12 | img_data = preprocess_input(x)
13 | classes = model.predict(img_data)
14 | 


--------------------------------------------------------------------------------
/Lungs.py:
--------------------------------------------------------------------------------
  1 | 
  2 | 
  3 | # -*- coding: utf-8 -*-
  4 | """
  5 | @author: Krish.Naik
  6 | """
  7 | 
  8 | from keras.layers import Input, Lambda, Dense, Flatten
  9 | from keras.models import Model
 10 | from keras.applications.vgg16 import VGG16
 11 | from keras.applications.vgg16 import preprocess_input
 12 | from keras.preprocessing import image
 13 | from keras.preprocessing.image import ImageDataGenerator
 14 | from keras.models import Sequential
 15 | import numpy as np
 16 | from glob import glob
 17 | import matplotlib.pyplot as plt
 18 | 
 19 | # re-size all the images to this
 20 | IMAGE_SIZE = [224, 224]
 21 | 
 22 | train_path = 'Datasets/train'
 23 | valid_path = 'Datasets/test'
 24 | 
 25 | # add preprocessing layer to the front of VGG
 26 | vgg = VGG16(input_shape=IMAGE_SIZE + [3], weights='imagenet', include_top=False)
 27 | 
 28 | # don't train existing weights
 29 | for layer in vgg.layers:
 30 |   layer.trainable = False
 31 |   
 32 | 
 33 |   
 34 |   # useful for getting number of classes
 35 | folders = glob('Datasets/train/*')
 36 |   
 37 | 
 38 | # our layers - you can add more if you want
 39 | x = Flatten()(vgg.output)
 40 | # x = Dense(1000, activation='relu')(x)
 41 | prediction = Dense(len(folders), activation='softmax')(x)
 42 | 
 43 | # create a model object
 44 | model = Model(inputs=vgg.input, outputs=prediction)
 45 | 
 46 | # view the structure of the model
 47 | model.summary()
 48 | 
 49 | # tell the model what cost and optimization method to use
 50 | model.compile(
 51 |   loss='categorical_crossentropy',
 52 |   optimizer='adam',
 53 |   metrics=['accuracy']
 54 | )
 55 | 
 56 | 
 57 | from keras.preprocessing.image import ImageDataGenerator
 58 | 
 59 | train_datagen = ImageDataGenerator(rescale = 1./255,
 60 |                                    shear_range = 0.2,
 61 |                                    zoom_range = 0.2,
 62 |                                    horizontal_flip = True)
 63 | 
 64 | test_datagen = ImageDataGenerator(rescale = 1./255)
 65 | 
 66 | training_set = train_datagen.flow_from_directory('Datasets/train',
 67 |                                                  target_size = (224, 224),
 68 |                                                  batch_size = 32,
 69 |                                                  class_mode = 'categorical')
 70 | 
 71 | test_set = test_datagen.flow_from_directory('Datasets/test',
 72 |                                             target_size = (224, 224),
 73 |                                             batch_size = 32,
 74 |                                             class_mode = 'categorical')
 75 | 
 76 | 
 77 | # fit the model
 78 | r = model.fit_generator(
 79 |   training_set,
 80 |   validation_data=test_set,
 81 |   epochs=5,
 82 |   steps_per_epoch=len(training_set),
 83 |   validation_steps=len(test_set)
 84 | )
 85 | # loss
 86 | plt.plot(r.history['loss'], label='train loss')
 87 | plt.plot(r.history['val_loss'], label='val loss')
 88 | plt.legend()
 89 | plt.show()
 90 | plt.savefig('LossVal_loss')
 91 | 
 92 | # accuracies
 93 | plt.plot(r.history['acc'], label='train acc')
 94 | plt.plot(r.history['val_acc'], label='val acc')
 95 | plt.legend()
 96 | plt.show()
 97 | plt.savefig('AccVal_acc')
 98 | 
 99 | import tensorflow as tf
100 | 
101 | from keras.models import load_model
102 | 
103 | model.save('model_vgg19.h5')
104 | 
105 | 


--------------------------------------------------------------------------------