docs: Add detailed comments to model training notebook

Added comprehensive comments explaining the CNN architecture, model compilation, training process, evaluation, and visualization steps to improve code documentation.

Author: arjun-ms

Model/Training.ipynb

@@ -15,6 +15,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Import pickle for loading preprocessed data\n",
     "import pickle"
    ]
   },
@@ -25,8 +26,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "X = pickle.load(open('X.pkl', 'rb'))\n",
-    "y = pickle.load(open('y.pkl', 'rb'))"
+    "# Load the preprocessed features and labels from pickle files\n",
+    "X = pickle.load(open('X.pkl', 'rb'))  # Load image features\n",
+    "y = pickle.load(open('y.pkl', 'rb'))  # Load corresponding labels"
    ]
   },
   {
@@ -36,6 +38,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Display the features array\n",
     "X"
    ]
   },
@@ -46,6 +49,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Display the labels array\n",
     "y"
    ]
   },
@@ -56,6 +60,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Normalize pixel values to range [0, 1] for better model training\n",
     "X = X/255"
    ]
   },
@@ -66,6 +71,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Display normalized features\n",
     "X"
    ]
   },
@@ -76,6 +82,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Check the current shape of the features array\n",
     "X.shape"
    ]
   },
@@ -88,6 +95,8 @@
    },
    "outputs": [],
    "source": [
+    "# Reshape data to include channel dimension for CNN input\n",
+    "# Shape: (samples, height, width, channels) where 1 = grayscale\n",
     "X = X.reshape(-1, 224, 224, 1)"
    ]
   },
@@ -98,6 +107,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Verify the new shape after reshaping\n",
     "X.shape"
    ]
   },
@@ -108,8 +118,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from keras.models import Sequential\n",
-    "from keras.layers import Conv2D, MaxPooling2D, Dense, Flatten, Dropout"
+    "# Import Keras libraries for building the CNN model\n",
+    "from keras.models import Sequential  # For creating sequential model\n",
+    "from keras.layers import Conv2D, MaxPooling2D, Dense, Flatten, Dropout  # Layer types"
    ]
   },
   {
@@ -119,7 +130,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# Initialising the CNN\n",
+    "# Initialize the Convolutional Neural Network\n",
     "model = Sequential()"
    ]
   },
@@ -130,27 +141,38 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# Step 1 - Convolution\n",
+    "# Step 1 - First Convolutional Layer\n",
+    "# 64 filters of size 3x3, ReLU activation function\n",
     "model.add(Conv2D(64,(3,3), activation='relu'))\n",
-    "# Step 2 - Pooling\n",
+    "\n",
+    "# Step 2 - First Pooling Layer\n",
+    "# 2x2 max pooling to reduce spatial dimensions\n",
     "model.add(MaxPooling2D((2,2)))\n",
     "\n",
-    "# Adding a second convolutional layer\n",
+    "# Adding a second convolutional block\n",
+    "# Another 64 filters of 3x3 for deeper feature extraction\n",
     "model.add(Conv2D(64,(3,3), activation='relu'))\n",
     "model.add(MaxPooling2D((2,2)))\n",
     "\n",
-    "# Adding a third convolutional layer\n",
+    "# Adding a third convolutional block\n",
+    "# Final convolutional layer for complex pattern recognition\n",
     "model.add(Conv2D(64,(3,3), activation='relu'))\n",
     "model.add(MaxPooling2D((2,2)))\n",
+    "\n",
+    "# Dropout layer to prevent overfitting by randomly dropping 40% of connections\n",
     "model.add(Dropout(0.4))\n",
     "\n",
     "# Step 3 - Flattening\n",
+    "# Convert 2D feature maps to 1D feature vector\n",
     "model.add(Flatten())\n",
     "\n",
-    "# Step 4 - Full Connection\n",
+    "# Step 4 - Full Connection (Hidden Layer)\n",
+    "# Dense layer with 128 neurons for learning complex patterns\n",
     "model.add(Dense(128, input_shape = X.shape[1:], activation = 'relu'))\n",
     "\n",
     "# Step 5 - Output Layer\n",
+    "# 3 neurons for 3 classes (Mammootty, Mohanlal, Unknown)\n",
+    "# Softmax activation for probability distribution across classes\n",
     "model.add(Dense(3, activation= 'softmax'))"
    ]
   },
@@ -160,10 +182,7 @@
    "id": "2f239525",
    "metadata": {},
    "outputs": [],
-   "source": [
-    "# Compiling the CNN\n",
-    "model.compile(optimizer = 'adam', loss='sparse_categorical_crossentropy', metrics = ['accuracy'])"
-   ]
+   "source": "# Compile the CNN model\n# Adam optimizer: adaptive learning rate optimization algorithm\n# sparse_categorical_crossentropy: loss function for multi-class classification with integer labels\n# accuracy: metric to monitor during training\nmodel.compile(optimizer = 'adam', loss='sparse_categorical_crossentropy', metrics = ['accuracy'])"
   },
   {
    "cell_type": "code",
@@ -173,92 +192,63 @@
     "scrolled": true
    },
    "outputs": [],
-   "source": [
-    "hist = model.fit(X, y, epochs=10, validation_split=0.2)"
-   ]
+   "source": "# Train the model\n# epochs=10: train for 10 complete passes through the dataset\n# validation_split=0.2: use 20% of data for validation during training\nhist = model.fit(X, y, epochs=10, validation_split=0.2)"
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "id": "cc34347a",
    "metadata": {},
    "outputs": [],
-   "source": [
-    "model.summary()"
-   ]
+   "source": "# Display model architecture summary\n# Shows layers, output shapes, and trainable parameters\nmodel.summary()"
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "id": "1f65bfca",
    "metadata": {},
    "outputs": [],
-   "source": [
-    "#to know accuracy of model\n",
-    "scores = model.evaluate(X,y,verbose=0)\n",
-    "print(\"Accuracy: %.2f%%\" % (scores[1]*100))"
-   ]
+   "source": "# Evaluate model accuracy on the training data\nscores = model.evaluate(X,y,verbose=0)\nprint(\"Accuracy: %.2f%%\" % (scores[1]*100))"
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "id": "582c9ac3",
    "metadata": {},
    "outputs": [],
-   "source": [
-    "X.shape"
-   ]
+   "source": "# Verify the shape of features array\nX.shape"
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "id": "466b56d3",
    "metadata": {},
    "outputs": [],
-   "source": [
-    "import matplotlib.pyplot as plt"
-   ]
+   "source": "# Import matplotlib for visualizing training metrics\nimport matplotlib.pyplot as plt"
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "id": "fb0ebddb",
    "metadata": {},
    "outputs": [],
-   "source": [
-    "fig = plt.figure()\n",
-    "plt.plot(hist.history['loss'],color='teal',label='loss')\n",
-    "plt.plot(hist.history['val_loss'],color='orange',label='val_loss')\n",
-    "plt.suptitle('Loss',fontsize=20)\n",
-    "plt.legend(loc=\"upper left\")\n",
-    "plt.show"
-   ]
+   "source": "# Plot training and validation loss over epochs\nfig = plt.figure()\nplt.plot(hist.history['loss'],color='teal',label='loss')  # Training loss\nplt.plot(hist.history['val_loss'],color='orange',label='val_loss')  # Validation loss\nplt.suptitle('Loss',fontsize=20)\nplt.legend(loc=\"upper left\")\nplt.show"
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "id": "36208bdd",
    "metadata": {},
    "outputs": [],
-   "source": [
-    "fig = plt.figure()\n",
-    "plt.plot(hist.history['accuracy'],color='teal',label='accuracy')\n",
-    "plt.plot(hist.history['val_accuracy'],color='orange',label='val_accuracy')\n",
-    "plt.suptitle('Accuracy',fontsize=20)\n",
-    "plt.legend(loc=\"upper left\")\n",
-    "plt.show"
-   ]
+   "source": "# Plot training and validation accuracy over epochs\nfig = plt.figure()\nplt.plot(hist.history['accuracy'],color='teal',label='accuracy')  # Training accuracy\nplt.plot(hist.history['val_accuracy'],color='orange',label='val_accuracy')  # Validation accuracy\nplt.suptitle('Accuracy',fontsize=20)\nplt.legend(loc=\"upper left\")\nplt.show"
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "id": "9418fe74",
    "metadata": {},
    "outputs": [],
-   "source": [
-    "# save the model\n",
-    "model.save('3-class-improved.h5')"
-   ]
+   "source": "# Save the trained model to disk\n# Saves model architecture, weights, and optimizer state\nmodel.save('3-class-improved.h5')"
   }
  ],
  "metadata": {
@@ -282,4 +272,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 5
-}
+}