tensorflow-uit2721dlca 0.1.0__tar.gz

tensorflow_uit2721dlca-0.1.0/PKG-INFO

@@ -0,0 +1,60 @@
+Metadata-Version: 2.4
+Name: tensorflow-uit2721dlca
+Version: 0.1.0
+Summary: Deep Learning assignment
+Requires-Python: <3.11,>=3.9
+Description-Content-Type: text/markdown
+Requires-Dist: tensorflow<2.10.0,>=2.9.0
+Requires-Dist: numpy<1.22.0,>=1.21.6
+Requires-Dist: matplotlib<3.6.0,>=3.5.2
+Requires-Dist: scikit-learn<2.0.0,>=1.0.0
+
+# tensorflow-uit2721dlca
+
+Deep Learning assignment package — UIT2721DLCA.
+
+## Requirements
+
+- Python 3.9 or 3.10 (TensorFlow 2.9 does not support Python 3.11+)
+
+## Installation
+
+`ash
+pip install tensorflow-uit2721dlca
+`
+
+## Usage
+
+Each module is a self-contained assignment exercise. Importing a module will
+execute its code (download data, train model, display plots).
+
+`python
+# Run exercise 1 — VGG16 image classification on Imagenette
+import tensorflow_uit2721dlca.ex1
+
+# Run exercise 2 — VGG16 feature maps and filters
+import tensorflow_uit2721dlca.ex2
+
+# Run exercise 3 — Transfer learning / fine-tuning
+import tensorflow_uit2721dlca.ex3
+
+# Run exercise 4 — Image captioning with CNN + LSTM
+import tensorflow_uit2721dlca.ex4
+
+# Run exercise 5 — Conditional GAN (text-to-image)
+import tensorflow_uit2721dlca.ex5
+
+# Run the prerequisites / environment check
+import tensorflow_uit2721dlca.prerequisites
+`
+
+> **Note:** The exercises download the Imagenette dataset (~100 MB) on first
+> run via 	f.keras.utils.get_file. The file is cached in ~/.keras/datasets/
+> and is not re-downloaded on subsequent runs.
+
+## Package name vs import name
+
+| Purpose | Name |
+|---|---|
+| pip install | 	ensorflow-uit2721dlca |
+| Python import | 	ensorflow_uit2721dlca |

tensorflow_uit2721dlca-0.1.0/README.md

@@ -0,0 +1,49 @@
+# tensorflow-uit2721dlca
+
+Deep Learning assignment package — UIT2721DLCA.
+
+## Requirements
+
+- Python 3.9 or 3.10 (TensorFlow 2.9 does not support Python 3.11+)
+
+## Installation
+
+`ash
+pip install tensorflow-uit2721dlca
+`
+
+## Usage
+
+Each module is a self-contained assignment exercise. Importing a module will
+execute its code (download data, train model, display plots).
+
+`python
+# Run exercise 1 — VGG16 image classification on Imagenette
+import tensorflow_uit2721dlca.ex1
+
+# Run exercise 2 — VGG16 feature maps and filters
+import tensorflow_uit2721dlca.ex2
+
+# Run exercise 3 — Transfer learning / fine-tuning
+import tensorflow_uit2721dlca.ex3
+
+# Run exercise 4 — Image captioning with CNN + LSTM
+import tensorflow_uit2721dlca.ex4
+
+# Run exercise 5 — Conditional GAN (text-to-image)
+import tensorflow_uit2721dlca.ex5
+
+# Run the prerequisites / environment check
+import tensorflow_uit2721dlca.prerequisites
+`
+
+> **Note:** The exercises download the Imagenette dataset (~100 MB) on first
+> run via 	f.keras.utils.get_file. The file is cached in ~/.keras/datasets/
+> and is not re-downloaded on subsequent runs.
+
+## Package name vs import name
+
+| Purpose | Name |
+|---|---|
+| pip install | 	ensorflow-uit2721dlca |
+| Python import | 	ensorflow_uit2721dlca |

tensorflow_uit2721dlca-0.1.0/pyproject.toml

@@ -0,0 +1,20 @@
+[build-system]
+requires = ["setuptools>=61"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "tensorflow-uit2721dlca"
+version = "0.1.0"
+description = "Deep Learning assignment"
+readme = "README.md"
+
+requires-python = ">=3.9,<3.11"
+dependencies = [
+    "tensorflow >=2.9.0,<2.10.0",
+    "numpy >=1.21.6,<1.22.0",
+    "matplotlib >=3.5.2,<3.6.0",
+    "scikit-learn >=1.0.0,<2.0.0",
+]
+
+[tool.setuptools.packages.find]
+where = ["src"]

tensorflow_uit2721dlca-0.1.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.1.0"

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca/ex1.py

@@ -0,0 +1,68 @@
+import tensorflow as tf
+import matplotlib.pyplot as plt
+
+from tensorflow.keras.applications import VGG16
+from tensorflow.keras.applications.vgg16 import preprocess_input, decode_predictions
+
+# Download Imagenette
+url = "https://s3.amazonaws.com/fast-ai-imageclas/imagenette2-160.tgz"
+
+archive = tf.keras.utils.get_file(
+    "imagenette2-160.tgz",
+    url,
+    untar=True
+)
+
+# Load validation images
+dataset = tf.keras.utils.image_dataset_from_directory(
+    archive + "/imagenette2-160/val",
+    image_size=(224, 224),
+    batch_size=5,
+    shuffle=True
+)
+
+print("Classes:", dataset.class_names)
+
+# Get 5 images
+images, labels = next(iter(dataset))
+
+print("Original shape:", images.shape)
+print("Preprocessed shape: (5, 224, 224, 3)")
+
+# Display images and original labels
+plt.figure(figsize=(12, 5))
+
+for i in range(5):
+    plt.subplot(1, 5, i + 1)
+    plt.imshow(images[i].numpy().astype("uint8"))
+    plt.title(dataset.class_names[labels[i]])
+    plt.axis("off")
+
+plt.show()
+
+# Load pretrained VGG16
+model = VGG16(weights="imagenet")
+
+# Preprocess images
+x = preprocess_input(images)
+
+# Show preprocessed representation
+print("Pixel value before preprocessing:", images[0][0][0])
+print("Pixel value after preprocessing :", x[0][0][0])
+
+# Predict
+predictions = model.predict(x)
+
+# Top 5 predictions
+results = decode_predictions(predictions, top=5)
+
+# Display predictions and compare with original labels
+for i in range(5):
+    print("\nImage", i + 1)
+    print("Original label:", dataset.class_names[labels[i]])
+    print("Top-5 predictions:")
+
+    for _, label, probability in results[i]:
+        print(label, ":", round(probability * 100, 2), "%")
+
+    print("Predicted label:", results[i][0][1])

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca/ex2.py

@@ -0,0 +1,71 @@
+import tensorflow as tf
+import matplotlib.pyplot as plt
+from tensorflow.keras.applications import VGG16
+from tensorflow.keras.applications.vgg16 import preprocess_input
+
+# Download Imagenette
+url = "https://s3.amazonaws.com/fast-ai-imageclas/imagenette2-160.tgz"
+
+archive = tf.keras.utils.get_file(
+    "imagenette2-160.tgz",
+    url,
+    untar=True
+)
+
+# Load validation images
+dataset = tf.keras.utils.image_dataset_from_directory(
+    archive + "/imagenette2-160/val",
+    image_size=(224, 224),
+    batch_size=5,
+    shuffle=True
+)
+
+images, labels = next(iter(dataset))
+
+# Load VGG16
+model = VGG16(weights="imagenet")
+
+# Take one image
+image = images[0:1]
+x = preprocess_input(image)
+
+# Select 3 layers
+layers = ["block1_conv1", "block2_conv2", "block5_conv3"]
+
+# Create model to get feature maps
+feature_model = tf.keras.Model(
+    inputs=model.input,
+    outputs=[model.get_layer(layer).output for layer in layers]
+)
+
+features = feature_model.predict(x)
+
+# Display feature maps
+for layer, feature in zip(layers, features):
+
+    print(layer, "shape:", feature.shape)
+
+    plt.figure(figsize=(10, 5))
+
+    for i in range(6):
+        plt.subplot(2, 3, i + 1)
+        plt.imshow(feature[0, :, :, i], cmap="viridis")
+        plt.axis("off")
+
+    plt.suptitle(layer)
+    plt.show()
+
+# Display filters from first convolutional layer
+filters = model.get_layer("block1_conv1").get_weights()[0]
+
+print("Filter shape:", filters.shape)
+
+plt.figure(figsize=(10, 5))
+
+for i in range(6):
+    plt.subplot(2, 3, i + 1)
+    plt.imshow(filters[:, :, :, i])
+    plt.axis("off")
+
+plt.suptitle("VGG16 Convolutional Filters")
+plt.show()

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca/ex3.py

@@ -0,0 +1,73 @@
+import tensorflow as tf
+import numpy as np
+
+from tensorflow.keras.applications import VGG16
+from tensorflow.keras.applications.vgg16 import preprocess_input
+from tensorflow.keras.layers import GlobalAveragePooling2D, Dense
+from tensorflow.keras.models import Sequential
+from sklearn.metrics import confusion_matrix, classification_report
+
+# Download Imagenette
+url = "https://s3.amazonaws.com/fast-ai-imageclas/imagenette2-160.tgz"
+
+archive = tf.keras.utils.get_file(
+    "imagenette2-160.tgz",
+    url,
+    untar=True
+)
+
+# Load only a small dataset
+train = tf.keras.utils.image_dataset_from_directory(
+    archive + "/imagenette2-160/train",
+    image_size=(224, 224),
+    batch_size=20,
+    shuffle=True
+).take(10)
+
+test = tf.keras.utils.image_dataset_from_directory(
+    archive + "/imagenette2-160/val",
+    image_size=(224, 224),
+    batch_size=20,
+    shuffle=True
+).take(3)
+
+# Preprocess
+train = train.map(lambda x, y: (preprocess_input(x), y))
+test = test.map(lambda x, y: (preprocess_input(x), y))
+
+# VGG16 feature extractor
+base = VGG16(weights="imagenet", include_top=False)
+print("Feature map shape:", base.output.shape)
+base.trainable = False
+
+# Classifier
+model = Sequential([
+    base,
+    GlobalAveragePooling2D(),
+    Dense(10, activation="softmax")
+])
+print("Feature vector size: 512")
+
+model.compile(
+    optimizer="adam",
+    loss="sparse_categorical_crossentropy",
+    metrics=["accuracy"]
+)
+
+# Train
+model.fit(train, epochs=1)
+
+# Test
+y_true = []
+y_pred = []
+
+for x, y in test:
+    pred = model.predict(x, verbose=0)
+    y_true.extend(y.numpy())
+    y_pred.extend(np.argmax(pred, axis=1))
+
+print("Confusion Matrix:")
+print(confusion_matrix(y_true, y_pred))
+
+print("\nAccuracy, Precision, Recall and F1:")
+print(classification_report(y_true, y_pred))

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca/ex4.py

@@ -0,0 +1,199 @@
+import tensorflow as tf
+import numpy as np
+import matplotlib.pyplot as plt
+
+from tensorflow.keras.applications import VGG16
+from tensorflow.keras.applications.vgg16 import preprocess_input
+from tensorflow.keras.preprocessing.text import Tokenizer
+from tensorflow.keras.preprocessing.sequence import pad_sequences
+from tensorflow.keras.layers import Input, Dense, LSTM, Embedding
+from tensorflow.keras.models import Model
+
+# Download Imagenette
+url = "https://s3.amazonaws.com/fast-ai-imageclas/imagenette2-160.tgz"
+
+archive = tf.keras.utils.get_file(
+    "imagenette2-160.tgz",
+    url,
+    untar=True
+)
+
+# Load 10 images
+dataset = tf.keras.utils.image_dataset_from_directory(
+    archive + "/imagenette2-160/val",
+    image_size=(224, 224),
+    batch_size=10,
+    shuffle=True
+)
+
+images, labels = next(iter(dataset))
+
+# ImageNet class names
+names = {
+    "n01440764": "fish",
+    "n02102040": "dog",
+    "n02979186": "cassette player",
+    "n03000684": "chainsaw",
+    "n03028079": "church",
+    "n03394916": "French horn",
+    "n03417042": "garbage truck",
+    "n03425413": "gas pump",
+    "n03445777": "golf ball",
+    "n03888257": "parachute"
+}
+
+# Create captions
+captions = []
+
+for label in labels:
+    name = names[dataset.class_names[label]]
+    captions.append("start a photo of a " + name + " end")
+
+print("Example:", captions[0])
+
+# Tokenize
+tokenizer = Tokenizer()
+tokenizer.fit_on_texts(captions)
+
+sequences = tokenizer.texts_to_sequences(captions)
+
+vocab_size = len(tokenizer.word_index) + 1
+max_length = max(len(x) for x in sequences)
+
+print("Vocabulary size:", vocab_size)
+
+# CNN feature extraction
+cnn = VGG16(
+    weights="imagenet",
+    include_top=False,
+    pooling="avg"
+)
+
+cnn.trainable = False
+
+features = cnn.predict(
+    preprocess_input(images),
+    verbose=0
+)
+
+print("Feature shape:", features.shape)
+
+# Create training examples
+X_image = []
+X_text = []
+Y = []
+
+for i in range(len(sequences)):
+
+    seq = sequences[i]
+
+    for j in range(1, len(seq)):
+
+        X_image.append(features[i])
+
+        X_text.append(
+            seq[:j]
+        )
+
+        Y.append(
+            seq[j]
+        )
+
+X_image = np.array(X_image)
+
+X_text = pad_sequences(
+    X_text,
+    maxlen=max_length,
+    padding="pre"
+)
+
+Y = np.array(Y)
+
+# LSTM decoder
+image_input = Input(shape=(512,))
+text_input = Input(shape=(max_length,))
+
+image = Dense(128, activation="tanh")(image_input)
+
+text = Embedding(
+    vocab_size,
+    128
+)(text_input)
+
+lstm = LSTM(128)(text, initial_state=[image, image])
+
+output = Dense(
+    vocab_size,
+    activation="softmax"
+)(lstm)
+
+model = Model(
+    [image_input, text_input],
+    output
+)
+
+model.compile(
+    optimizer="adam",
+    loss="sparse_categorical_crossentropy",
+    metrics=["accuracy"]
+)
+
+# Train
+model.fit(
+    [X_image, X_text],
+    Y,
+    epochs=100,
+    verbose=0
+)
+
+print("Training complete")
+
+# Generate caption
+feature = features[0:1]
+
+words = ["start"]
+
+for i in range(max_length):
+
+    sequence = tokenizer.texts_to_sequences(
+        [" ".join(words)]
+    )
+
+    sequence = pad_sequences(
+        sequence,
+        maxlen=max_length,
+        padding="pre"
+    )
+
+    prediction = model.predict(
+        [feature, sequence],
+        verbose=0
+    )
+
+    word_id = np.argmax(prediction[0])
+
+    word = tokenizer.index_word.get(
+        word_id,
+        ""
+    )
+
+    if word == "end" or word == "":
+        break
+
+    words.append(word)
+
+caption = " ".join(words[1:])
+
+print("Generated Caption:", caption)
+
+# Display image
+plt.imshow(
+    images[0].numpy().astype("uint8")
+)
+
+plt.title(
+    caption
+)
+
+plt.axis("off")
+plt.show()

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca/ex5.py

@@ -0,0 +1,70 @@
+import tensorflow as tf
+import numpy as np
+import matplotlib.pyplot as plt
+from tensorflow.keras import layers
+
+# Text representation
+texts = ["dog", "tiger", "bird", "truck", "church"]
+text = np.eye(5)
+
+# Generator: noise + text -> image
+generator = tf.keras.Sequential([
+    layers.Input(shape=(105,)),
+    layers.Dense(128, activation="relu"),
+    layers.Dense(64 * 64 * 3, activation="sigmoid"),
+    layers.Reshape((64, 64, 3))
+])
+
+# Discriminator: image + text -> real/fake
+image = layers.Input(shape=(64, 64, 3))
+condition = layers.Input(shape=(5,))
+
+x = layers.Flatten()(image)
+x = layers.Concatenate()([x, condition])
+x = layers.Dense(64, activation="relu")(x)
+output = layers.Dense(1, activation="sigmoid")(x)
+
+discriminator = tf.keras.Model([image, condition], output)
+
+discriminator.compile(
+    optimizer="adam",
+    loss="binary_crossentropy"
+)
+
+# Random noise + text
+noise = np.random.normal(0, 1, (5, 100))
+fake_images = generator.predict(
+    np.concatenate([noise, text], axis=1),
+    verbose=0
+)
+
+# Train discriminator
+real_images = np.random.random((5, 64, 64, 3))
+
+discriminator.train_on_batch(
+    [real_images, text],
+    np.ones((5, 1))
+)
+
+discriminator.train_on_batch(
+    [fake_images, text],
+    np.zeros((5, 1))
+)
+
+print("Generated image shape:", fake_images.shape)
+print("Discriminator output:",
+      discriminator.predict([fake_images, text], verbose=0))
+
+# Unseen text
+new_text = np.array([[1, 0, 0, 0, 0]])
+new_noise = np.random.normal(0, 1, (1, 100))
+
+new_image = generator.predict(
+    np.concatenate([new_noise, new_text], axis=1),
+    verbose=0
+)
+
+plt.imshow(new_image[0])
+plt.title("Generated image for unseen text")
+plt.axis("off")
+plt.show()

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca/prerequisites.py

@@ -0,0 +1,87 @@
+import tensorflow as tf
+import numpy as np
+import matplotlib.pyplot as plt
+
+from tensorflow.keras.applications import VGG16
+from tensorflow.keras.applications.vgg16 import preprocess_input
+from tensorflow.keras.applications.vgg16 import decode_predictions
+
+print("TensorFlow version:", tf.__version__)
+
+
+#============================
+
+url = "https://s3.amazonaws.com/fast-ai-imageclas/imagenette2-160.tgz"
+
+archive = tf.keras.utils.get_file(
+    "imagenette2-160.tgz",
+    url,
+    untar=True
+)
+
+print(archive)
+
+#=================================
+
+dataset = tf.keras.utils.image_dataset_from_directory(
+    archive + "/imagenette2-160/val",
+    image_size=(224, 224),
+    batch_size=5,
+    shuffle=True
+)
+
+print(dataset.class_names)
+
+
+#================================
+
+images, labels = next(iter(dataset))
+
+print("Original shape:", images.shape)
+print("Preprocessed shape: (5, 224, 224, 3)")
+
+
+#=====================================
+
+plt.figure(figsize=(12, 5))
+
+for i in range(5):
+    plt.subplot(1, 5, i + 1)
+    plt.imshow(images[i].numpy().astype("uint8"))
+    plt.title(dataset.class_names[labels[i]])
+    plt.axis("off")
+
+plt.show()
+
+
+#======================================
+
+model = VGG16(weights="imagenet")
+
+# Preprocess images
+x = preprocess_input(images)
+
+# Show preprocessed representation
+print("Pixel value before preprocessing:", images[0][0][0])
+print("Pixel value after preprocessing :", x[0][0][0])
+
+
+
+#======================================
+
+
+# Predict
+predictions = model.predict(x)
+
+results = decode_predictions(predictions, top=5)
+
+# Display predictions and compare with original labels
+for i in range(5):
+    print("\nImage", i + 1)
+    print("Original label:", dataset.class_names[labels[i]])
+    print("Top-5 predictions:")
+
+    for _, label, probability in results[i]:
+        print(label, ":", round(probability * 100, 2), "%")
+
+    print("Predicted label:", results[i][0][1])

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca.egg-info/PKG-INFO

@@ -0,0 +1,60 @@
+Metadata-Version: 2.4
+Name: tensorflow-uit2721dlca
+Version: 0.1.0
+Summary: Deep Learning assignment
+Requires-Python: <3.11,>=3.9
+Description-Content-Type: text/markdown
+Requires-Dist: tensorflow<2.10.0,>=2.9.0
+Requires-Dist: numpy<1.22.0,>=1.21.6
+Requires-Dist: matplotlib<3.6.0,>=3.5.2
+Requires-Dist: scikit-learn<2.0.0,>=1.0.0
+
+# tensorflow-uit2721dlca
+
+Deep Learning assignment package — UIT2721DLCA.
+
+## Requirements
+
+- Python 3.9 or 3.10 (TensorFlow 2.9 does not support Python 3.11+)
+
+## Installation
+
+`ash
+pip install tensorflow-uit2721dlca
+`
+
+## Usage
+
+Each module is a self-contained assignment exercise. Importing a module will
+execute its code (download data, train model, display plots).
+
+`python
+# Run exercise 1 — VGG16 image classification on Imagenette
+import tensorflow_uit2721dlca.ex1
+
+# Run exercise 2 — VGG16 feature maps and filters
+import tensorflow_uit2721dlca.ex2
+
+# Run exercise 3 — Transfer learning / fine-tuning
+import tensorflow_uit2721dlca.ex3
+
+# Run exercise 4 — Image captioning with CNN + LSTM
+import tensorflow_uit2721dlca.ex4
+
+# Run exercise 5 — Conditional GAN (text-to-image)
+import tensorflow_uit2721dlca.ex5
+
+# Run the prerequisites / environment check
+import tensorflow_uit2721dlca.prerequisites
+`
+
+> **Note:** The exercises download the Imagenette dataset (~100 MB) on first
+> run via 	f.keras.utils.get_file. The file is cached in ~/.keras/datasets/
+> and is not re-downloaded on subsequent runs.
+
+## Package name vs import name
+
+| Purpose | Name |
+|---|---|
+| pip install | 	ensorflow-uit2721dlca |
+| Python import | 	ensorflow_uit2721dlca |

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca.egg-info/SOURCES.txt

@@ -0,0 +1,14 @@
+README.md
+pyproject.toml
+src/tensorflow_uit2721dlca/__init__.py
+src/tensorflow_uit2721dlca/ex1.py
+src/tensorflow_uit2721dlca/ex2.py
+src/tensorflow_uit2721dlca/ex3.py
+src/tensorflow_uit2721dlca/ex4.py
+src/tensorflow_uit2721dlca/ex5.py
+src/tensorflow_uit2721dlca/prerequisites.py
+src/tensorflow_uit2721dlca.egg-info/PKG-INFO
+src/tensorflow_uit2721dlca.egg-info/SOURCES.txt
+src/tensorflow_uit2721dlca.egg-info/dependency_links.txt
+src/tensorflow_uit2721dlca.egg-info/requires.txt
+src/tensorflow_uit2721dlca.egg-info/top_level.txt

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca.egg-info/requires.txt

@@ -0,0 +1,4 @@
+tensorflow<2.10.0,>=2.9.0
+numpy<1.22.0,>=1.21.6
+matplotlib<3.6.0,>=3.5.2
+scikit-learn<2.0.0,>=1.0.0

tensorflow_uit2721dlca-0.1.0/src/tensorflow_uit2721dlca.egg-info/top_level.txt

@@ -0,0 +1 @@
+tensorflow_uit2721dlca