Upload 6 files

autoencoders_MNIST_denoiser.py

@@ -0,0 +1,126 @@
+# Autoencoder usages for denoising MNIST dataset.
+# Learning how to use autoencoder for a denoising task.
+# This will later be used in my project for denoising sensor data of MOS for gas sensing.
+
+# YouTube Resource Credits: https://www.youtube.com/watch?v=Sm54KXD-L1k
+
+# We are going to encode an image and then decode it to see if the image is denoised, in which the bottleneck layer will loss some information,
+# which will be used to learn the features of the image.
+
+# Noise Reduction is a common problem, in which autoencoders are used to solve this problem by reconstructing the image.
+
+# We will be using Keras and Tensorflow for this task.
+from keras.datasets import mnist
+from tensorflow.keras.layers import Conv2D, MaxPooling2D, UpSampling2D
+from tensorflow.keras.models import Sequential
+import numpy as np
+import random, cv2
+from matplotlib import pyplot as plt
+
+
+def sp_noise(image,prob):
+    '''
+    Add salt and pepper noise to image
+    prob: Probability of the noise
+    '''
+    output = np.zeros(image.shape,np.uint8)
+    thres = 1 - prob
+    for i in range(image.shape[0]):
+        for j in range(image.shape[1]):
+            rdn = random.random()
+            if rdn < prob:
+                output[i][j] = 0
+            elif rdn > thres:
+                output[i][j] = 255
+            else:
+                output[i][j] = image[i][j]
+    return output
+
+(x_train, _), (x_test, _) = mnist.load_data()  # We are not interested in the labels.  28x28x1 images.
+
+# Same as normal face constructor, we simply test based on the MNIST and (transformed) blurred dataset
+
+# Convert to 0-1 range
+x_train = x_train.astype('float32') / 255.0
+x_test = x_test.astype('float32') / 255.0
+
+# Reshape the dataset
+x_train = x_train.reshape((len(x_train), 28, 28, 1))
+x_test = x_test.reshape((len(x_test), 28, 28, 1))
+
+
+# Add noise to the dataset
+
+# Guassin noise factor.
+noise_factor = 0.5
+x_train_noisy = x_train + noise_factor * np.random.normal(loc=0.0, scale=1.0, size=x_train.shape)
+x_test_noisy = x_test + noise_factor * np.random.normal(loc=0.0, scale=1.0, size=x_test.shape)
+
+# Clip the dataset to 0-1 range
+x_train_noisy = np.clip(x_train_noisy, 0., 1.)
+x_test_noisy = np.clip(x_test_noisy, 0., 1.)
+
+#Displaying images with noise
+# plt.figure(figsize=(20, 2))
+# for i in range(1,10):
+#     ax = plt.subplot(1, 10, i)
+#     plt.imshow(x_test_noisy[i].reshape(28, 28), cmap="binary")
+# plt.show()
+
+SIZE = 28
+DIMENSIONS = 1
+model = Sequential()
+# Encoder Layer
+model.add(Conv2D(32, (3, 3), activation='relu', padding='same', input_shape=(SIZE, SIZE, DIMENSIONS)))
+model.add(MaxPooling2D((2, 2), padding='same')) # Shrinking
+model.add(Conv2D(64, (3, 3), activation='relu', padding='same'))
+model.add(MaxPooling2D((2, 2), padding='same')) # Shrinking
+model.add(Conv2D(128, (3, 3), activation='relu', padding='same'))
+model.add(MaxPooling2D((2, 2), padding='same')) # Bottleneck
+model.add(Conv2D(256, (3, 3), activation='relu', padding='same'))
+model.add(MaxPooling2D((2, 2), padding='same')) # Bottleneck
+model.add(Conv2D(512, (3, 3), activation='relu', padding='same'))
+
+# Middle Layer
+model.add(MaxPooling2D((2, 2), padding='same')) # Shrinking
+
+# Decoder Layer
+model.add(Conv2D(512, (3, 3), activation='relu', padding='same'))
+model.add(UpSampling2D((2, 2)))  # Expanding
+model.add(Conv2D(256, (3, 3), activation='relu', padding='same'))
+model.add(UpSampling2D((2, 2)))  # Expanding
+model.add(Conv2D(128, (3, 3), activation='relu', padding='same'))
+model.add(UpSampling2D((2, 2)))  # Expanding
+model.add(Conv2D(64, (3, 3), activation='relu', padding='same'))
+model.add(UpSampling2D((2, 2)))  # Expanding
+model.add(Conv2D(32, (3, 3), activation='relu', padding='same'))
+model.add(UpSampling2D((2, 2)))  # Expanding
+
+# Output Layer
+model.add(Conv2D(DIMENSIONS, (5, 5), activation='relu'))
+
+model.compile(optimizer='adam', loss='mean_squared_error', metrics=['accuracy'])
+
+model.summary()
+model.fit(x=x_train_noisy, y=x_train, epochs=3, batch_size=256, shuffle=True, validation_data=(x_test_noisy, x_test))
+
+
+model.evaluate(x_test_noisy, x_test)
+
+
+# Save model and show model outputs.
+model.save('denoising_autoencoder.model')
+
+no_noise_img = model.predict(x_test_noisy)
+
+plt.figure(figsize=(40, 4))
+for i in range(10):
+	# display original
+	ax = plt.subplot(3, 20, i + 1)
+	plt.imshow(x_test_noisy[i].reshape(28, 28), cmap="binary")
+
+	# display reconstructed (after noise removed) image
+	ax = plt.subplot(3, 20, 40 + i + 1)
+	plt.imshow(no_noise_img[i].reshape(28, 28), cmap="binary")
+
+plt.waitforbuttonpress()

autoencoders_face_reconstructions.py

@@ -0,0 +1,136 @@
+# Learning the basic structure of an autoencoder.
+import tensorflow as tf
+print("Num GPUs Available: ", len(tf.config.list_physical_devices('GPU')))
+
+
+from tensorflow.keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D
+from tensorflow.keras.models import Sequential, Model
+from keras.utils.image_utils import img_to_array, load_img
+from matplotlib.pyplot import imshow
+from PIL import Image, ImageDraw
+import numpy as np
+import cv2
+
+np.random.seed(342)  # Seed for reproducibility, we want to see the same results every time we run the code,
+# reproducibility.
+
+# Image Preprocessing
+# Convert to an image numpy array: (float32)
+# Batches
+# Normalise the image.
+profile_img = cv2.imread('profile.PNG')
+profile_img = cv2.cvtColor(profile_img, cv2.COLOR_BGR2RGB)  # convert to RGB.
+print(profile_img.shape)  # (28, 28, 3)
+
+reshaped = cv2.resize(profile_img, (256, 256))
+print(reshaped.shape)  # (256, 256, 3)
+img_array = img_to_array(reshaped).astype('float32') / 255.  # Normalise the image.
+# Need 1 batch of the image.
+img_array = np.expand_dims(img_array, axis=0)
+
+# cv2.imshow('Image', reshaped)
+# cv2.waitKey(0)
+
+# Get the image dimensions at batch x height x width x channels
+# Roughly 1x28x28x3 dims.
+print(img_array.shape)  # (1, 256, 256, 3)
+
+SIZE = 256
+model = Sequential()
+# Encoder Layer
+model.add(Conv2D(32, (3, 3), activation='relu', padding='same', input_shape=(SIZE, SIZE, 3)))
+model.add(MaxPooling2D((2, 2), padding='same')) # Shrinking
+model.add(Conv2D(64, (3, 3), activation='relu', padding='same'))
+model.add(MaxPooling2D((2, 2), padding='same')) # Shrinking
+model.add(Conv2D(128, (3, 3), activation='relu', padding='same'))
+model.add(MaxPooling2D((2, 2), padding='same')) # Bottleneck
+model.add(Conv2D(256, (3, 3), activation='relu', padding='same'))
+model.add(MaxPooling2D((2, 2), padding='same')) # Bottleneck
+model.add(Conv2D(512, (3, 3), activation='relu', padding='same'))
+
+# Middle Layer
+model.add(MaxPooling2D((2, 2), padding='same')) # Shrinking
+
+# Decoder Layer
+model.add(Conv2D(512, (3, 3), activation='relu', padding='same'))
+model.add(UpSampling2D((2, 2)))  # Expanding
+model.add(Conv2D(256, (3, 3), activation='relu', padding='same'))
+model.add(UpSampling2D((2, 2)))  # Expanding
+model.add(Conv2D(128, (3, 3), activation='relu', padding='same'))
+model.add(UpSampling2D((2, 2)))  # Expanding
+model.add(Conv2D(64, (3, 3), activation='relu', padding='same'))
+model.add(UpSampling2D((2, 2)))  # Expanding
+model.add(Conv2D(32, (3, 3), activation='relu', padding='same'))
+model.add(UpSampling2D((2, 2)))  # Expanding
+
+# Output Layer
+model.add(Conv2D(3, (3, 3), activation='relu', padding='same'))
+
+model.compile(optimizer='adam', loss='mean_squared_error', metrics=['accuracy'])
+
+
+
+
+if __name__ == '__main__':
+	model.summary()
+
+	# Image Gif Generator based on keras callbacks.
+	model_outputs = []
+
+
+	def on_epoch_end(epoch, logs):
+		model_outputs.append(model.predict(img_array))
+
+
+	# Callbacks
+	predict_callback = tf.keras.callbacks.LambdaCallback(on_epoch_end=on_epoch_end)
+
+	# I realised Keras is really easy to use, and I am loving it.
+	# Training the model
+	model.fit(x=img_array, y=img_array, epochs=1000, batch_size=20, shuffle=True, callbacks=[predict_callback])
+
+
+	predict = model.predict(x=img_array)  # We hope to get similar to the original image.
+	cv2.imshow("output", predict[0].reshape(SIZE, SIZE, 3))
+	cv2.waitKey(0)
+
+	print("Creating a gif/video from the model outputs.")
+	# Make a gif creator from an array of given images.
+	print(len(model_outputs))  # 50
+
+	images = [(output[0] * 255).astype('uint8') for output in model_outputs]
+	try:
+		images = [(output[0] * 255).astype('uint8') for output in model_outputs]
+
+		# Specify the video file name and codec
+		video_filename = 'model_outputs_5.4.mp4'
+		fourcc = cv2.VideoWriter_fourcc(*'mp4v')  # Use appropriate codec for your system
+
+		# Get the height and width from the first image
+		height, width, _ = images[0].shape
+
+		# Create a VideoWriter object
+		fps = 30  # Adjust the frames per second (FPS) as needed
+		video = cv2.VideoWriter(video_filename, fourcc, fps, (width, height))
+
+		# Write each image to the video
+		for image in images:
+			# Convert BRG to RGB
+			rgb_image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+			video.write(rgb_image)
+
+		# Release the VideoWriter
+		video.release()
+	except Exception as e:
+		print(e)
+		images = [Image.fromarray((output[0] * 255).astype('uint8')) for output in model_outputs]
+		images[0].save('model_outputs_5.4m_param.gif', save_all=True, append_images=images[1:], optimize=False, duration=20,
+		               loop=0)
+	#
+
+	# # Save the images as a GIF
+	#
+
+	images = [Image.fromarray((output[0] * 255).astype('uint8')) for output in model_outputs]
+	images[0].save('model_outputs_5.4m_param.gif', save_all=True, append_images=images[1:], optimize=False, duration=20,
+		               loop=0)

denoising_autoencoder.model/keras_metadata.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68231aeb20103b779df4e60a2b53505d605cea8000cf0d18540dffbbbb393641
+size 45912

denoising_autoencoder.model/saved_model.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:216047dc3ab6b1cb38c3d93874e16a02f506d04f4fed5da2108c8fb67b262859
+size 348938

denoising_autoencoder.model/variables/variables.data-00000-of-00001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd07a3d1b98010a29c044ee1019dbc2394c481c2d8a0432ece60d03d6a287c0e
+size 65972683

denoising_autoencoder.model/variables/variables.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0129990d23ee787363b7d7209534166ec9a34a1fe87d19ec0192c230e5df4b5
+size 5371