megaage-asian

MobileNet_and_DenseNet/CA_D.py

@@ -0,0 +1,73 @@
+import numpy as np
+import sys
+import logging
+import csv
+import os
+os.environ['TF_CPP_MIN_LOG_LEVEL']='2'
+from TYY_utils import mk_dir, load_data_npz
+from TYY_model import TYY_DenseNet_reg
+
+def MAE(a,b):
+  mae = np.sum(np.absolute(a-b))
+  mae/=len(b)
+  return mae
+
+'''''''''''''''''''''''''''''''''''''''''''''
+  file name
+'''''''''''''''''''''''''''''''''''''''''''''
+test_file = sys.argv[1]
+netType = int(sys.argv[2])
+
+logging.debug("Loading testing data...")
+image2, age2, image_size = load_data_npz(test_file)
+
+if netType == 3:
+    N_densenet = 3
+    depth_densenet = 3*N_densenet+4
+    model_file = 'megaage_models/DenseNet/batch_size_50/densenet_reg_%d_64/densenet_reg_%d_64.h5'%(depth_densenet, depth_densenet)
+    model = TYY_DenseNet_reg(image_size,depth_densenet)()
+    mk_dir('Results_csv')
+    save_name = 'Results_csv/densenet_reg_%d_%d.csv' % (depth_densenet, image_size)
+
+elif netType == 4:
+    N_densenet = 5
+    depth_densenet = 3*N_densenet+4
+    model_file = 'megaage_models/DenseNet/batch_size_50/densenet_reg_%d_64/densenet_reg_%d_64.h5'%(depth_densenet, depth_densenet)
+    model = TYY_DenseNet_reg(image_size,depth_densenet)()
+    mk_dir('Results_csv')
+    save_name = 'Results_csv/densenet_reg_%d_%d.csv' % (depth_densenet, image_size)
+
+
+'''''''''''''''''''''''''''''''''''''''''''''
+  load data
+'''''''''''''''''''''''''''''''''''''''''''''
+logging.debug("Loading model file...")
+model.load_weights(model_file)
+
+age_p=model.predict(image2)
+
+'''''''''''''''''''''''''''''''''''''''''''''
+  prediction
+'''''''''''''''''''''''''''''''''''''''''''''
+pred=[['MAE'],[str(MAE(age2,age_p[:,0]))],['CA3','CA5'],['0','0'],['ID','age','age_p','error']]
+CA3=0
+CA5=0
+for i in range(0,len(image2)):
+  error=np.absolute(age2[i]-age_p[i,0])
+  if error<=3:
+    CA3+=1
+  if error<=5:
+    CA5+=1
+  temp = [str(i), str(age2[i]), str(age_p[i,0]), str(error)]
+  pred.append(temp)
+
+CA3/=len(image2)
+CA5/=len(image2)
+pred[3]=[str(CA3),str(CA5)]
+
+print('CA3: ',CA3,'\nCA5: ',CA5)
+
+f=open(save_name,'w')
+w=csv.writer(f)
+w.writerows(pred)
+f.close

MobileNet_and_DenseNet/CA_M.py

@@ -0,0 +1,74 @@
+import numpy as np
+import sys
+import logging
+import csv
+import os
+os.environ['TF_CPP_MIN_LOG_LEVEL']='2'
+
+from TYY_utils import mk_dir, load_data_npz
+from TYY_model import TYY_MobileNet_reg
+
+def MAE(a,b):
+  mae = np.sum(np.absolute(a-b))
+  mae/=len(b)
+  return mae
+
+'''''''''''''''''''''''''''''''''''''''''''''
+  file name
+'''''''''''''''''''''''''''''''''''''''''''''
+test_file = sys.argv[1]
+netType = int(sys.argv[2])
+
+logging.debug("Loading testing data...")
+image2, age2, image_size = load_data_npz(test_file)
+
+if netType == 1:
+    alpha = 0.25
+    model_file = 'megaage_models/MobileNet/batch_size_50/mobilenet_reg_0.25_64/mobilenet_reg_0.25_64.h5'
+    model = TYY_MobileNet_reg(image_size,alpha)()
+    mk_dir('Results_csv')
+    save_name = 'Results_csv/mobilenet_reg_%s_%d.csv' % (alpha, image_size)
+
+elif netType == 2:
+    alpha = 0.5
+    model_file = 'megaage_models/MobileNet/batch_size_50/mobilenet_reg_0.5_64/mobilenet_reg_0.5_64.h5'
+    model = TYY_MobileNet_reg(image_size,alpha)()
+    mk_dir('Results_csv')
+    save_name = 'Results_csv/mobilenet_reg_%s_%d.csv' % (alpha, image_size)
+
+
+'''''''''''''''''''''''''''''''''''''''''''''
+  load data
+'''''''''''''''''''''''''''''''''''''''''''''
+logging.debug("Loading model file...")
+model.load_weights(model_file)
+
+age_p=model.predict(image2,batch_size=len(image2))
+
+'''''''''''''''''''''''''''''''''''''''''''''
+  prediction
+'''''''''''''''''''''''''''''''''''''''''''''
+pred=[['MAE'],[str(MAE(age2,age_p[:,0]))],['ID','age','age_p','error']]
+pred.append(['CA3','CA5'])
+pred.append(['0','0'])
+CA3=0
+CA5=0
+for i in range(0,len(image2)):
+  error=np.absolute(age2[i]-age_p[i,0])
+  if error<=3:
+    CA3+=1
+  if error<=5:
+    CA5+=1
+  temp = [str(i), str(age2[i]), str(age_p[i,0]), str(error)]
+  pred.append(temp)
+
+CA3/=len(image2)
+CA5/=len(image2)
+pred[4]=[str(CA3),str(CA5)]
+
+print('CA3: ',CA3,'\nCA5: ',CA5)
+
+f=open(save_name,'w')
+w=csv.writer(f)
+w.writerows(pred)
+f.close

MobileNet_and_DenseNet/TYY_callbacks.py

@@ -0,0 +1,36 @@
+import keras
+from sklearn.metrics import roc_auc_score
+import sys
+import matplotlib.pyplot as plt
+from keras.models import Model
+import numpy as np
+from keras import backend as K
+
+
+class DecayLearningRate(keras.callbacks.Callback):
+	def __init__(self, startEpoch):
+		self.startEpoch = startEpoch
+
+	def on_train_begin(self, logs={}):
+		return
+	def on_train_end(self, logs={}):
+		return
+
+	def on_epoch_begin(self, epoch, logs={}):
+		if epoch in self.startEpoch:
+			if epoch == 0:
+				ratio = 1
+			else:
+				ratio = 0.1
+			LR = K.get_value(self.model.optimizer.lr)
+			K.set_value(self.model.optimizer.lr,LR*ratio)
+		return
+
+	def on_epoch_end(self, epoch, logs={}):
+		return
+
+	def on_batch_begin(self, batch, logs={}):
+		return
+
+	def on_batch_end(self, batch, logs={}):
+		return

MobileNet_and_DenseNet/TYY_generators.py

@@ -0,0 +1,102 @@
+import keras
+import numpy as np
+import sys
+from scipy import misc
+import tensorflow as tf
+
+
+def random_crop(x,dn):
+    dx = np.random.randint(dn,size=1)[0]
+    dy = np.random.randint(dn,size=1)[0]
+    w = x.shape[0]
+    h = x.shape[1]
+    out = x[0+dx:w-(dn-dx),0+dy:h-(dn-dy),:]
+    out = misc.imresize(out, (w,h), interp='nearest')
+    return out
+
+def augment_data(images):
+    for i in range(0,images.shape[0]):
+
+        if np.random.random() > 0.5:
+            images[i] = images[i][:,::-1]
+        """
+        if np.random.random() > 0.5:
+            images[i] = random_crop(images[i],4)
+        """
+        if np.random.random() > 0.75:
+            images[i] = tf.contrib.keras.preprocessing.image.random_rotation(images[i], 20, row_axis=0, col_axis=1, channel_axis=2)
+        if np.random.random() > 0.75:
+            images[i] = tf.contrib.keras.preprocessing.image.random_shear(images[i], 0.2, row_axis=0, col_axis=1, channel_axis=2)
+        if np.random.random() > 0.75:
+            images[i] = tf.contrib.keras.preprocessing.image.random_shift(images[i], 0.2, 0.2, row_axis=0, col_axis=1, channel_axis=2)
+        if np.random.random() > 0.75:
+            images[i] = tf.contrib.keras.preprocessing.image.random_zoom(images[i], [0.8,1.2], row_axis=0, col_axis=1, channel_axis=2)
+
+    return images
+
+
+def data_generator_reg(X,Y,batch_size):
+
+    while True:
+        idxs = np.random.permutation(len(X))
+        X = X[idxs]
+        Y = Y[idxs]
+        p,q = [],[]
+        for i in range(len(X)):
+            p.append(X[i])
+            q.append(Y[i])
+            if len(p) == batch_size:
+                yield augment_data(np.array(p)),np.array(q)
+                p,q = [],[]
+        if p:
+            yield augment_data(np.array(p)),np.array(q)
+            p,q = [],[]
+
+def data_generator_dex(X,Y,batch_size):
+
+    Y1 = Y[0]
+    Y2 = Y[1]
+
+    while True:
+        idxs = np.random.permutation(len(X))
+        X = X[idxs]
+        Y1 = Y1[idxs]
+        Y2 = Y2[idxs]
+        p,q1,q2 = [],[],[]
+        for i in range(len(X)):
+            p.append(X[i])
+            q1.append(Y1[i])
+            q2.append(Y2[i])
+            if len(p) == batch_size:
+                yield augment_data(np.array(p)),[np.array(q1),np.array(q2)]
+                p,q1,q2 = [],[],[]
+        if p:
+            yield augment_data(np.array(p)),[np.array(q1),np.array(q2)]
+            p,q1,q2 = [],[],[]
+
+def data_generator_dex_centerloss(X,Y,batch_size):
+    X1 = X[0]
+    X2 = X[1]
+    Y1 = Y[0]
+    Y2 = Y[1]
+    Y3 = Y[2]
+    while True:
+        idxs = np.random.permutation(len(X1))
+        X1 = X1[idxs] #images
+        X2 = X2[idxs] #labels for center loss
+        Y1 = Y1[idxs]
+        Y2 = Y2[idxs]
+        Y3 = Y3[idxs]
+        p1,p2,q1,q2,q3 = [],[],[],[],[]
+        for i in range(len(X1)):
+            p1.append(X1[i])
+            p2.append(X2[i])
+            q1.append(Y1[i])
+            q2.append(Y2[i])
+            q3.append(Y3[i])
+            if len(p1) == batch_size:
+                yield [augment_data(np.array(p1)),np.array(p2)],[np.array(q1),np.array(q2),np.array(q3)]
+                p1,p2,q1,q2,q3 = [],[],[],[],[]
+        if p1:
+            yield [augment_data(np.array(p1)),np.array(p2)],[np.array(q1),np.array(q2),np.array(q3)]
+            p1,p2,q1,q2,q3 = [],[],[],[],[]

MobileNet_and_DenseNet/TYY_model.py

@@ -0,0 +1,86 @@
+# This code is imported from the following project: https://github.com/asmith26/wide_resnets_keras
+
+import logging
+import sys
+import numpy as np
+from keras.models import Model
+from keras.layers import Input, Activation, add, Dense, Flatten, Dropout, Multiply, Embedding, Lambda, Add, Concatenate, Activation
+from keras.layers.convolutional import Conv2D, AveragePooling2D, MaxPooling2D
+from keras.layers.normalization import BatchNormalization
+from keras.regularizers import l2
+from keras import backend as K
+from keras.optimizers import SGD,Adam
+from keras.applications.mobilenet import MobileNet
+from densenet import *
+from keras.utils import plot_model
+
+sys.setrecursionlimit(2 ** 20)
+np.random.seed(2 ** 10)
+
+
+class TYY_MobileNet_reg:
+    def __init__(self, image_size, alpha):
+
+
+        if K.image_dim_ordering() == "th":
+            logging.debug("image_dim_ordering = 'th'")
+            self._channel_axis = 1
+            self._input_shape = (3, image_size, image_size)
+        else:
+            logging.debug("image_dim_ordering = 'tf'")
+            self._channel_axis = -1
+            self._input_shape = (image_size, image_size, 3)
+        self.alpha = alpha
+
+#    def create_model(self):
+    def __call__(self):
+        logging.debug("Creating model...")
+
+        inputs = Input(shape=self._input_shape)
+        model_mobilenet = MobileNet(input_shape=self._input_shape, alpha=self.alpha, depth_multiplier=1, dropout=1e-3, include_top=False, weights=None, input_tensor=None, pooling=None)
+        x = model_mobilenet(inputs)
+        #flatten = Flatten()(x)
+
+        feat_a = Conv2D(20,(1,1),activation='relu')(x)
+        feat_a = Flatten()(feat_a)
+        feat_a = Dropout(0.2)(feat_a)
+        feat_a = Dense(32,activation='relu',name='feat_a')(feat_a)
+
+        pred_a = Dense(1,name='pred_a')(feat_a)
+        model = Model(inputs=inputs, outputs=[pred_a])
+
+
+        return model
+
+
+class TYY_DenseNet_reg:
+    def __init__(self, image_size, depth):
+
+        if K.image_dim_ordering() == "th":
+            logging.debug("image_dim_ordering = 'th'")
+            self._channel_axis = 1
+            self._input_shape = (3, image_size, image_size)
+        else:
+            logging.debug("image_dim_ordering = 'tf'")
+            self._channel_axis = -1
+            self._input_shape = (image_size, image_size, 3)
+        self.depth = depth
+
+#    def create_model(self):
+    def __call__(self):
+        logging.debug("Creating model...")
+
+        inputs = Input(shape=self._input_shape)
+        model_densenet = DenseNet(input_shape=self._input_shape, depth=self.depth, include_top=False, weights=None, input_tensor=None)
+        flatten = model_densenet(inputs)
+
+        feat_a = Dense(128,activation='relu')(flatten)
+        feat_a = Dropout(0.2)(feat_a)
+        feat_a = Dense(32,activation='relu',name='feat_a')(feat_a)
+
+        pred_a = Dense(1,name='pred_a')(feat_a)
+        model = Model(inputs=inputs, outputs=[pred_a])
+
+        return model
+
+