Merge pull request #221 from chawlaj100/master

Added Resizing.py to resize images

Author: master-fury

Image_Processing/src/cropping/Resizing.py

@@ -0,0 +1,19 @@
+"""
+This is small script used for resizing any image. Resizing is useful whenever we have an image dataset having
+images of different resolutions. Any model would accept a fixed sized image and so we need to resize the image.
+
+OpenCV provides a direct implementation of resizing the image.
+There are 3 arguments in resize()
+1. The image to be resized
+2. The desired width and height in a tuple.
+3. The kind of interpolation. (Choose from AREA, CUBIC, LINEAR, and NEAREST)
+"""
+import cv2
+dimensions = input("Enter width and height : ").split()
+width = int(dimensions[0])
+height = int(dimensions[1])
+
+image = cv2.imread("lenna.png")
+resized_image = cv2.resize(image , (width, height), interpolation=cv2.INTER_AREA)
+cv2.imshow("Resized Image", resized_image)
+cv2.waitKey(0)

Machine_Learning/src/MNIST_using_CNN/mnist.py

@@ -0,0 +1,69 @@
+"""
+This file describes implementation of a basic CNN made using Tensorflow and trained and tested with MNIST dataset.
+Initially we have all the required dependencies imported.
+
+create_model() describes the CNN architecture used for training a model
+
+plot() is used to plot the images along with their predicted/true labels.
+
+"""
+import numpy as np
+import tensorflow as tf
+import matplotlib.pyplot as plt
+import cv2
+
+# CNN Model 
+def create_model():
+    model = tf.keras.models.Sequential([
+        tf.keras.layers.Conv2D(16, (3,3), activation="relu", input_shape=(28,28,1)),
+        tf.keras.layers.MaxPooling2D((2,2)),
+        tf.keras.layers.Conv2D(32, (2,2), activation="relu"),
+        tf.keras.layers.MaxPooling2D((2,2)),
+        tf.keras.layers.Flatten(),
+        tf.keras.layers.Dense(64, activation="relu"),
+        tf.keras.layers.Dropout(0.3),
+        tf.keras.layers.Dense(10, activation="softmax")
+    ])
+    return model
+
+# Plot images with labels
+def plot(subset_x, subset_y):
+    for i in range(len(subset_y)):
+    plt.figure(figsize=(10,20))
+    ax = plt.subplot(len(subset_y)/2,2,i+1)
+    img = subset_x[i]
+    ax.imshow(img)
+    plt.show()
+    print("LABEL : ",subset_y[i])
+    print()
+
+# Load Mnist dataset from tensorflow datasets
+(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()
+
+# Normalize the pixel values and convert the lables to one-hot vectors
+x_train = x_train/255.
+x_test = x_test/255.
+x_train = x_train.reshape((-1,28,28,1))
+x_test = x_test.reshape((-1,28,28,1))
+y_train = tf.one_hot(y_train, depth=10)
+y_test = tf.one_hot(y_test, depth=10)
+
+# Create and compile the Model
+my_model = create_model()
+my_model.compile(optimizer = tf.keras.optimizers.Adam(lr=0.001), loss="categorical_crossentropy", metrics=["accuracy"])
+
+# Fit the created model and store the information for plotting of loss and accuracy curves
+history = my_model.fit(x_train, y_train, validation_split=0.1, epochs=5, batch_size=32, shuffle=True, verbose=1)
+
+# Evaluate the model using the test data
+loss, acc = my_model.evaluate(x_test, y_test)
+
+# Make Predictions
+y_predictions = my_model.predict_classes(x_test)
+
+subset_x = x_test[0:10]
+subset_y = y_predictions[0:10]
+
+# Used to plot the predicted labels along with images.
+plot(subset_x, subset_y)
+