Merge pull request #172 from jeffin07/master

implementaion of VGG16

Author: ionicc

Machine_Learning/src/VGG16/README.md

@@ -0,0 +1,7 @@
+This is the implemetaion of the VGG architecture(16 layers).I've used mnist dataset to train the model for simplicity<br/>
+
+The paper can be found in [VGG](https://arxiv.org/pdf/1409.1556.pdf) <br/>
+
+And the model architecture for VGG is given below the implemetation is for the configuration D(16 layers)<br/>
+
+![](imgs/vggnet_table1.png) 

Machine_Learning/src/VGG16/imgs/vggnet_table1.png

@@ -0,0 +1,119 @@
+import tensorflow as tf 
+from tensorflow.keras.models import Model
+from tensorflow.keras.losses import MSE
+from tensorflow.keras.optimizers import Adam
+import numpy as np
+
+class Vgg16(tf.keras.Model):
+	def __init__(self,output_nodes):
+		super(Vgg16, self).__init__()
+		# layers needed
+		self.conv1_1 = tf.keras.layers.Conv2D(
+					input_shape=[None,28,28,1],filters=64, kernel_size=3,
+					padding="same", activation="relu")
+		self.conv1_2 = tf.keras.layers.Conv2D(
+					filters=64, kernel_size=3,
+					padding="same", activation="relu")
+		self.conv2_1 = tf.keras.layers.Conv2D(
+					filters=128, kernel_size=3,
+					padding="same", activation="relu")
+		self.conv2_2 = tf.keras.layers.Conv2D(
+					filters=128, kernel_size=3,
+					padding="same", activation="relu")
+		self.conv3_1 = tf.keras.layers.Conv2D(
+					filters=256,kernel_size=3,
+					padding="same", activation="relu")
+		self.conv3_2 = tf.keras.layers.Conv2D(
+					filters=256,kernel_size=3,
+					padding="same", activation="relu")
+		self.conv3_3 = tf.keras.layers.Conv2D(
+					filters=256,kernel_size=3,
+					padding="same", activation="relu")
+		
+		self.conv4_1 = tf.keras.layers.Conv2D(
+					filters=512, kernel_size=3,
+					padding="same", activation="relu")
+		self.conv4_2 = tf.keras.layers.Conv2D(
+					filters=512, kernel_size=3,
+					padding="same", activation="relu")
+		self.conv4_3 = tf.keras.layers.Conv2D(
+					filters=512, kernel_size=3,
+					padding="same", activation="relu")
+		self.conv5_1 = tf.keras.layers.Conv2D(
+					filters=512, kernel_size=3,
+					padding="same", activation="relu")
+		self.conv5_2 = tf.keras.layers.Conv2D(
+					filters=512, kernel_size=3,
+					padding="same", activation="relu")
+		self.conv5_3 = tf.keras.layers.Conv2D(
+					filters=512, kernel_size=3,
+					padding="same", activation="relu")
+		self.dense1_1 = tf.keras.layers.Dense(
+					units=4096, activation="relu")
+		self.dense1_2 = tf.keras.layers.Dense(
+					units=4096, activation="relu")
+		self.dense2 = tf.keras.layers.Dense(
+					units=output_nodes, activation="softmax")
+		self.maxPool = tf.keras.layers.MaxPool2D(
+					pool_size=2, strides=2, padding="same")
+		self.flatten = tf.keras.layers.Flatten()
+
+	def call(self,input):
+		# ops 
+		x = self.conv1_1(input)
+		x = self.conv1_2(x)
+		x = self.maxPool(x)
+		x = self.conv2_1(x)
+		x = self.conv2_2(x)
+		x = self.maxPool(x)
+		x = self.conv3_1(x)
+		x = self.conv3_2(x)
+		x = self.conv3_3(x)
+		x = self.maxPool(x)
+		x = self.conv4_1(x)
+		x = self.conv4_2(x)
+		x = self.conv4_3(x)
+		x = self.maxPool(x)
+		x = self.conv5_1(x)
+		x = self.conv5_2(x)
+		x = self.conv5_3(x)
+		x = self.maxPool(x)
+		x = self.flatten(x)
+		x = self.dense1_1(x)
+		x = self.dense1_2(x)
+		x = self.dense2(x)
+		return x
+
+
+network = Vgg16(10)
+# network.compile(optimizer="adam", loss="sparse_categorical_crossentropy", metrics=["accuracy"])
+
+mnist = tf.keras.datasets.mnist
+
+(x_train, y_train),(x_test, y_test) = mnist.load_data()
+x_train = (x_train.reshape(-1, 28, 28, 1) / 255).astype(np.float32)
+# x_train, x_test = x_train / 255.0, x_test / 255.0
+y_train = np.eye(10)[y_train].astype(np.float32)
+x_train = x_train[:100]
+y_train = y_train[:100]
+
+
+
+input_layer = tf.keras.layers.Input(shape=(28,28,1))
+output_layer = network(input_layer)
+training_model = Model(inputs=input_layer,outputs=output_layer)
+optim = Adam()
+
+for i in range(100):
+	with tf.GradientTape(watch_accessed_variables=False) as tape:
+		tape.watch(training_model.trainable_variables)
+		preds = training_model(x_train)
+		loss = MSE(preds, y_train)
+		cost = tf.reduce_mean(loss)
+		grads = tape.gradient(loss, training_model.trainable_variables)
+		optim.apply_gradients(zip(grads, training_model.trainable_variables))
+	print(cost)
+
+
+
+