load and update

Author: fjadidi2001

CM_GAN_Jan5.ipynb

@@ -5,7 +5,7 @@
     "colab": {
       "provenance": [],
       "gpuType": "T4",
-      "authorship_tag": "ABX9TyNJ/r/NM/jYEo7auYrkb7pZ",
+      "authorship_tag": "ABX9TyPcTRx4Ut5mkhdIsmwBpSfx",
       "include_colab_link": true
     },
     "kernelspec": {
@@ -184,6 +184,121 @@
         "id": "lT70RwhUq8O-"
       }
     },
+    {
+      "cell_type": "code",
+      "source": [
+        "import os\n",
+        "import numpy as np\n",
+        "import tensorflow as tf\n",
+        "from tensorflow.keras.preprocessing.image import load_img, img_to_array\n",
+        "from skimage.draw import random_shapes\n",
+        "from sklearn.model_selection import train_test_split\n",
+        "import matplotlib.pyplot as plt\n",
+        "\n",
+        "# Parameters\n",
+        "DATA_DIR = \"./places2\"  # Path to your dataset\n",
+        "IMG_SIZE = (128, 128)  # Target image size\n",
+        "BATCH_SIZE = 32\n",
+        "MASK_TYPE = \"random\"  # Options: \"random\", \"rectangle\"\n",
+        "\n",
+        "# Generate a binary mask\n",
+        "def generate_mask(img_size, mask_type=\"random\"):\n",
+        "    if mask_type == \"random\":\n",
+        "        # Generate random irregular mask\n",
+        "        mask, _ = random_shapes(img_size, max_shapes=5, min_size=50, max_size=100, multichannel=False)\n",
+        "        mask = (mask == 255).astype(np.float32)  # Convert to binary mask\n",
+        "    elif mask_type == \"rectangle\":\n",
+        "        # Generate rectangular mask\n",
+        "        mask = np.ones(img_size, dtype=np.float32)\n",
+        "        x1, y1 = np.random.randint(0, img_size[0] // 2), np.random.randint(0, img_size[1] // 2)\n",
+        "        x2, y2 = np.random.randint(x1, img_size[0]), np.random.randint(y1, img_size[1])\n",
+        "        mask[x1:x2, y1:y2] = 0\n",
+        "    return mask\n",
+        "\n",
+        "# Load and preprocess a single image\n",
+        "def load_and_preprocess_image(img_path, img_size):\n",
+        "    img = load_img(img_path, target_size=img_size)\n",
+        "    img = img_to_array(img) / 255.0  # Normalize to [0, 1]\n",
+        "    return img\n",
+        "\n",
+        "# Create a TensorFlow Dataset\n",
+        "def create_dataset(data_dir, img_size, batch_size, mask_type=\"random\"):\n",
+        "    # Get list of image paths\n",
+        "    image_paths = [os.path.join(data_dir, img_name) for img_name in os.listdir(data_dir)\n",
+        "                   if img_name.endswith(('.jpg', '.png', '.jpeg'))]\n",
+        "\n",
+        "    if not image_paths:\n",
+        "        raise ValueError(f\"No images found in {data_dir}. Please check the dataset path.\")\n",
+        "\n",
+        "    # Split into training and validation sets\n",
+        "    train_paths, val_paths = train_test_split(image_paths, test_size=0.2, random_state=42)\n",
+        "\n",
+        "    # Function to load and preprocess images and masks\n",
+        "    def process_image(img_path):\n",
+        "        # Load and preprocess image\n",
+        "        img = tf.numpy_function(load_and_preprocess_image, [img_path, img_size], tf.float32)\n",
+        "        img.set_shape(img_size + (3,))  # Set shape explicitly\n",
+        "\n",
+        "        # Generate mask\n",
+        "        mask = tf.numpy_function(generate_mask, [img_size, mask_type], tf.float32)\n",
+        "        mask.set_shape(img_size)  # Set shape explicitly\n",
+        "\n",
+        "        # Apply mask to image\n",
+        "        masked_img = img * tf.expand_dims(mask, axis=-1)\n",
+        "\n",
+        "        return masked_img, img, mask\n",
+        "\n",
+        "    # Create TensorFlow Dataset\n",
+        "    train_dataset = tf.data.Dataset.from_tensor_slices(train_paths)\n",
+        "    train_dataset = train_dataset.map(process_image, num_parallel_calls=tf.data.AUTOTUNE)\n",
+        "    train_dataset = train_dataset.batch(batch_size).prefetch(tf.data.AUTOTUNE)\n",
+        "\n",
+        "    val_dataset = tf.data.Dataset.from_tensor_slices(val_paths)\n",
+        "    val_dataset = val_dataset.map(process_image, num_parallel_calls=tf.data.AUTOTUNE)\n",
+        "    val_dataset = val_dataset.batch(batch_size).prefetch(tf.data.AUTOTUNE)\n",
+        "\n",
+        "    return train_dataset, val_dataset\n",
+        "\n",
+        "# Create the dataset\n",
+        "try:\n",
+        "    train_dataset, val_dataset = create_dataset(DATA_DIR, IMG_SIZE, BATCH_SIZE, MASK_TYPE)\n",
+        "    print(\"Dataset created successfully.\")\n",
+        "except Exception as e:\n",
+        "    print(f\"Error creating dataset: {e}\")\n",
+        "\n",
+        "# Visualize a sample batch\n",
+        "def visualize_batch(dataset):\n",
+        "    for masked_images, original_images, masks in dataset.take(1):\n",
+        "        plt.figure(figsize=(10, 5))\n",
+        "        for i in range(3):  # Display 3 samples\n",
+        "            plt.subplot(3, 3, i * 3 + 1)\n",
+        "            plt.title(\"Masked Image\")\n",
+        "            plt.imshow(masked_images[i])\n",
+        "            plt.axis(\"off\")\n",
+        "\n",
+        "            plt.subplot(3, 3, i * 3 + 2)\n",
+        "            plt.title(\"Original Image\")\n",
+        "            plt.imshow(original_images[i])\n",
+        "            plt.axis(\"off\")\n",
+        "\n",
+        "            plt.subplot(3, 3, i * 3 + 3)\n",
+        "            plt.title(\"Mask\")\n",
+        "            plt.imshow(masks[i], cmap='gray')\n",
+        "            plt.axis(\"off\")\n",
+        "        plt.show()\n",
+        "\n",
+        "# Visualize a batch from the training dataset\n",
+        "if 'train_dataset' in locals():\n",
+        "    visualize_batch(train_dataset)\n",
+        "else:\n",
+        "    print(\"Dataset not available for visualization.\")"
+      ],
+      "metadata": {
+        "id": "IB80Tsl8GEDY"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
     {
       "cell_type": "markdown",
       "source": [