{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"scrolled": true
},
"outputs": [],
"source": [
"import numpy as np\n",
"from tensorflow.keras.layers import BatchNormalization, Conv2D, MaxPooling2D\n",
"from tensorflow.keras.preprocessing.image import load_img, img_to_array\n",
"from sklearn.metrics import mean_squared_error\n",
"from tensorflow.keras.initializers import RandomNormal\n",
"from tensorflow.keras.applications.vgg16 import VGG16\n",
"from tensorflow.keras.optimizers import SGD\n",
"from tensorflow.keras.models import Model, Sequential\n",
"from tensorflow.keras import backend as K\n",
"from tensorflow.keras.models import model_from_json\n",
"from matplotlib import cm as CM\n",
"import matplotlib.pyplot as plt\n",
"import tensorflow as tf\n",
"from tqdm import tqdm\n",
"import scipy.io as io\n",
"from PIL import Image\n",
"import PIL\n",
"import h5py\n",
"import os\n",
"import glob\n",
"import cv2\n",
"import random\n",
"import math\n",
"import sys"
]
},
{
"cell_type": "code",
"execution_count": 18,
"metadata": {},
"outputs": [],
"source": [
"K.clear_session()\n",
"root = 'DATA'"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"part_A_train = os.path.join(root, 'part_A_final/train_data', 'images')\n",
"part_A_test = os.path.join(root, 'part_A_final/test_data', 'images')\n",
"part_B_train = os.path.join(root, 'part_B_final/train_data', 'images')\n",
"part_B_test = os.path.join(root, 'part_B_final/test_data', 'images')\n",
"temp = 'test_images'\n",
"path_sets = [part_A_train]"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Total images: 300\n"
]
}
],
"source": [
"img_paths = []\n",
"for path in path_sets:\n",
" for img_path in glob.glob(os.path.join(path, '*.jpg')):\n",
" img_paths.append(str(img_path))\n",
"print(\"Total images:\", len(img_paths))"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"def create_img(path):\n",
" \"\"\"Function to load, normalize and return image\"\"\"\n",
" im = Image.open(path).convert('RGB')\n",
" im = np.array(im)\n",
" im = im / 255.0\n",
" \n",
" im[:, :, 0] = (im[:, :, 0] - 0.485) / 0.229\n",
" im[:, :, 1] = (im[:, :, 1] - 0.456) / 0.224\n",
" im[:, :, 2] = (im[:, :, 2] - 0.406) / 0.225\n",
" \n",
" return im\n",
"\n",
"\n",
"def get_input(path):\n",
" path = path[0]\n",
" img = create_img(path)\n",
" return img\n",
"\n",
"\n",
"def get_output(path):\n",
" \"\"\"Import and resize target density map\"\"\"\n",
" gt_file = h5py.File(path, 'r')\n",
" target = np.asarray(gt_file['density'])\n",
" \n",
" img = cv2.resize(target, (int(target.shape[1]/8), int(target.shape[0]/8)), \n",
" interpolation=cv2.INTER_CUBIC) * 64\n",
" img = np.expand_dims(img, axis=2) # Changed from axis=3 to axis=2\n",
" \n",
" return img\n",
"\n",
"\n",
"\n",
"def preprocess_input(image, target):\n",
" \"\"\"Crop image and target (optional data augmentation)\"\"\"\n",
" crop_size = (int(image.shape[0]/2), int(image.shape[1]/2))\n",
" \n",
" if random.randint(0, 9) <= -1:\n",
" dx = int(random.randint(0, 1) * image.shape[0] * 1./2)\n",
" dy = int(random.randint(0, 1) * image.shape[1] * 1./2)\n",
" else:\n",
" dx = int(random.random() * image.shape[0] * 1./2)\n",
" dy = int(random.random() * image.shape[1] * 1./2)\n",
" \n",
" img = image[dx:crop_size[0]+dx, dy:crop_size[1]+dy]\n",
" target_aug = target[dx:crop_size[0]+dx, dy:crop_size[1]+dy]\n",
" \n",
" return (img, target_aug)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"def image_generator(files, batch_size=1):\n",
" \"\"\"Image data generator\"\"\"\n",
" import numpy as np # Add this line\n",
" while True:\n",
" input_path = np.random.choice(a=files, size=1)\n",
" \n",
" inputt = get_input(input_path)\n",
" # Match the path from preprocessing\n",
" output_path = input_path[0].replace('.jpg', '.h5').replace('images', 'ground')\n",
" \n",
" if not os.path.exists(output_path):\n",
" print(f\"File not found: {output_path}\")\n",
" continue\n",
" \n",
" output = get_output(output_path)\n",
" \n",
" batch_x = np.expand_dims(inputt, axis=0)\n",
" batch_y = np.expand_dims(output, axis=0)\n",
" \n",
" yield (batch_x, batch_y)\n"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"def save_mod(model, str1, str2):\n",
" \"\"\"Save model weights and architecture\"\"\"\n",
" model.save_weights(str1)\n",
" model_json = model.to_json()\n",
" \n",
" with open(str2, \"w\") as json_file:\n",
" json_file.write(model_json)"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [],
"source": [
"def init_weights_vgg(model):\n",
" \"\"\"Initialize VGG16 pretrained weights\"\"\"\n",
" # Use built-in VGG16 with ImageNet weights\n",
" vgg = VGG16(weights='imagenet', include_top=False)\n",
" \n",
" vgg_weights = []\n",
" for layer in vgg.layers:\n",
" if 'conv' in layer.name:\n",
" vgg_weights.append(layer.get_weights())\n",
" \n",
" offset = 0\n",
" i = 0\n",
" while i < 10:\n",
" if 'conv' in model.layers[i+offset].name:\n",
" model.layers[i+offset].set_weights(vgg_weights[i])\n",
" i = i + 1\n",
" else:\n",
" offset = offset + 1\n",
" \n",
" return model\n"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [],
"source": [
"def euclidean_distance_loss(y_true, y_pred):\n",
" \"\"\"Euclidean distance as a measure of loss (Loss function)\"\"\"\n",
" return tf.sqrt(tf.reduce_sum(tf.square(y_pred - y_true), axis=-1))\n"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [],
"source": [
"# def CrowdNet():\n",
"# \"\"\"Neural network model: VGG + Conv\"\"\"\n",
"# # Variable Input Size\n",
"# rows = None\n",
"# cols = None\n",
" \n",
"# # Batch Normalisation option\n",
"# batch_norm = 0\n",
"# kernel = (3, 3)\n",
"# init = RandomNormal(stddev=0.01)\n",
"# model = Sequential()\n",
" \n",
"# # Custom VGG\n",
"# if batch_norm:\n",
"# model.add(Conv2D(64, kernel_size=kernel, input_shape=(rows, cols, 3), \n",
"# activation='relu', padding='same'))\n",
"# model.add(BatchNormalization())\n",
"# model.add(Conv2D(64, kernel_size=kernel, activation='relu', padding='same'))\n",
"# model.add(BatchNormalization())\n",
"# model.add(MaxPooling2D(strides=2))\n",
"# model.add(Conv2D(128, kernel_size=kernel, activation='relu', padding='same'))\n",
"# model.add(BatchNormalization())\n",
"# model.add(Conv2D(128, kernel_size=kernel, activation='relu', padding='same'))\n",
"# model.add(BatchNormalization())\n",
"# model.add(MaxPooling2D(strides=2))\n",
"# model.add(Conv2D(256, kernel_size=kernel, activation='relu', padding='same'))\n",
"# model.add(BatchNormalization())\n",
"# model.add(Conv2D(256, kernel_size=kernel, activation='relu', padding='same'))\n",
"# model.add(BatchNormalization())\n",
"# model.add(Conv2D(256, kernel_size=kernel, activation='relu', padding='same'))\n",
"# model.add(BatchNormalization())\n",
"# model.add(MaxPooling2D(strides=2))\n",
"# model.add(Conv2D(512, kernel_size=kernel, activation='relu', padding='same'))\n",
"# model.add(BatchNormalization())\n",
"# model.add(Conv2D(512, kernel_size=kernel, activation='relu', padding='same'))\n",
"# model.add(BatchNormalization())\n",
"# model.add(Conv2D(512, kernel_size=kernel, activation='relu', padding='same'))\n",
"# model.add(BatchNormalization())\n",
"# else:\n",
"# model.add(Conv2D(64, kernel_size=kernel, activation='relu', padding='same',\n",
"# input_shape=(rows, cols, 3), kernel_initializer=init))\n",
"# model.add(Conv2D(64, kernel_size=kernel, activation='relu', padding='same', \n",
"# kernel_initializer=init))\n",
"# model.add(MaxPooling2D(strides=2))\n",
"# model.add(Conv2D(128, kernel_size=kernel, activation='relu', padding='same', \n",
"# kernel_initializer=init))\n",
"# model.add(Conv2D(128, kernel_size=kernel, activation='relu', padding='same', \n",
"# kernel_initializer=init))\n",
"# model.add(MaxPooling2D(strides=2))\n",
"# model.add(Conv2D(256, kernel_size=kernel, activation='relu', padding='same', \n",
"# kernel_initializer=init))\n",
"# model.add(Conv2D(256, kernel_size=kernel, activation='relu', padding='same', \n",
"# kernel_initializer=init))\n",
"# model.add(Conv2D(256, kernel_size=kernel, activation='relu', padding='same', \n",
"# kernel_initializer=init))\n",
"# model.add(MaxPooling2D(strides=2))\n",
"# model.add(Conv2D(512, kernel_size=kernel, activation='relu', padding='same', \n",
"# kernel_initializer=init))\n",
"# model.add(Conv2D(512, kernel_size=kernel, activation='relu', padding='same', \n",
"# kernel_initializer=init))\n",
"# model.add(Conv2D(512, kernel_size=kernel, activation='relu', padding='same', \n",
"# kernel_initializer=init))\n",
" \n",
"# # Dilated Conv2D layers\n",
"# model.add(Conv2D(512, (3, 3), activation='relu', dilation_rate=2, \n",
"# kernel_initializer=init, padding='same'))\n",
"# model.add(Conv2D(512, (3, 3), activation='relu', dilation_rate=2, \n",
"# kernel_initializer=init, padding='same'))\n",
"# model.add(Conv2D(512, (3, 3), activation='relu', dilation_rate=2, \n",
"# kernel_initializer=init, padding='same'))\n",
"# model.add(Conv2D(256, (3, 3), activation='relu', dilation_rate=2, \n",
"# kernel_initializer=init, padding='same'))\n",
"# model.add(Conv2D(128, (3, 3), activation='relu', dilation_rate=2, \n",
"# kernel_initializer=init, padding='same'))\n",
"# model.add(Conv2D(64, (3, 3), activation='relu', dilation_rate=2, \n",
"# kernel_initializer=init, padding='same'))\n",
"# model.add(Conv2D(1, (1, 1), activation='relu', dilation_rate=1, \n",
"# kernel_initializer=init, padding='same'))\n",
" \n",
"# # Fixed: Use learning_rate instead of lr, removed decay\n",
"\n",
"# sgd = SGD(learning_rate=1e-7, momentum=0.95)\n",
"# model.compile(optimizer=sgd, loss='mse', metrics=['mse']) # Use 'mse' instead\n",
" \n",
"# model = init_weights_vgg(model)\n",
"# return model\n",
"\n",
"def CrowdNet():\n",
" \"\"\"Neural network model: VGG + Conv\"\"\"\n",
" rows = None\n",
" cols = None\n",
" batch_norm = 0\n",
" kernel = (3, 3)\n",
" init = RandomNormal(stddev=0.01)\n",
" model = Sequential()\n",
" \n",
" # VGG layers (without batch norm)\n",
" model.add(Conv2D(64, kernel_size=kernel, activation='relu', padding='same',\n",
" input_shape=(rows, cols, 3), kernel_initializer=init))\n",
" model.add(Conv2D(64, kernel_size=kernel, activation='relu', padding='same', kernel_initializer=init))\n",
" model.add(MaxPooling2D(strides=2))\n",
" model.add(Conv2D(128, kernel_size=kernel, activation='relu', padding='same', kernel_initializer=init))\n",
" model.add(Conv2D(128, kernel_size=kernel, activation='relu', padding='same', kernel_initializer=init))\n",
" model.add(MaxPooling2D(strides=2))\n",
" model.add(Conv2D(256, kernel_size=kernel, activation='relu', padding='same', kernel_initializer=init))\n",
" model.add(Conv2D(256, kernel_size=kernel, activation='relu', padding='same', kernel_initializer=init))\n",
" model.add(Conv2D(256, kernel_size=kernel, activation='relu', padding='same', kernel_initializer=init))\n",
" model.add(MaxPooling2D(strides=2))\n",
" model.add(Conv2D(512, kernel_size=kernel, activation='relu', padding='same', kernel_initializer=init))\n",
" model.add(Conv2D(512, kernel_size=kernel, activation='relu', padding='same', kernel_initializer=init))\n",
" model.add(Conv2D(512, kernel_size=kernel, activation='relu', padding='same', kernel_initializer=init))\n",
" \n",
" # Dilated Conv layers\n",
" model.add(Conv2D(512, (3, 3), activation='relu', dilation_rate=2, kernel_initializer=init, padding='same'))\n",
" model.add(Conv2D(512, (3, 3), activation='relu', dilation_rate=2, kernel_initializer=init, padding='same'))\n",
" model.add(Conv2D(512, (3, 3), activation='relu', dilation_rate=2, kernel_initializer=init, padding='same'))\n",
" model.add(Conv2D(256, (3, 3), activation='relu', dilation_rate=2, kernel_initializer=init, padding='same'))\n",
" model.add(Conv2D(128, (3, 3), activation='relu', dilation_rate=2, kernel_initializer=init, padding='same'))\n",
" model.add(Conv2D(64, (3, 3), activation='relu', dilation_rate=2, kernel_initializer=init, padding='same'))\n",
" model.add(Conv2D(1, (1, 1), activation='relu', dilation_rate=1, kernel_initializer=init, padding='same'))\n",
" \n",
" sgd = SGD(learning_rate=1e-7, momentum=0.95)\n",
" model.compile(optimizer=sgd, loss='mse', metrics=['mse'], run_eagerly=True) # Added run_eagerly=True\n",
" \n",
" model = init_weights_vgg(model)\n",
" return model\n",
"\n",
" "
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Building model...\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"c:\\Users\\adity\\anaconda3\\Lib\\site-packages\\keras\\src\\layers\\convolutional\\base_conv.py:113: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.\n",
" super().__init__(activity_regularizer=activity_regularizer, **kwargs)\n"
]
}
],
"source": [
"# Build model\n",
"print(\"Building model...\")\n",
"model = CrowdNet()\n"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"Model: \"sequential\"\n",
"\n"
],
"text/plain": [
"\u001b[1mModel: \"sequential\"\u001b[0m\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
"┃ Layer (type) ┃ Output Shape ┃ Param # ┃\n",
"┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
"│ conv2d (Conv2D) │ (None, None, None, 64) │ 1,792 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_1 (Conv2D) │ (None, None, None, 64) │ 36,928 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ max_pooling2d (MaxPooling2D) │ (None, None, None, 64) │ 0 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_2 (Conv2D) │ (None, None, None, │ 73,856 │\n",
"│ │ 128) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_3 (Conv2D) │ (None, None, None, │ 147,584 │\n",
"│ │ 128) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ max_pooling2d_1 (MaxPooling2D) │ (None, None, None, │ 0 │\n",
"│ │ 128) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_4 (Conv2D) │ (None, None, None, │ 295,168 │\n",
"│ │ 256) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_5 (Conv2D) │ (None, None, None, │ 590,080 │\n",
"│ │ 256) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_6 (Conv2D) │ (None, None, None, │ 590,080 │\n",
"│ │ 256) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ max_pooling2d_2 (MaxPooling2D) │ (None, None, None, │ 0 │\n",
"│ │ 256) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_7 (Conv2D) │ (None, None, None, │ 1,180,160 │\n",
"│ │ 512) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_8 (Conv2D) │ (None, None, None, │ 2,359,808 │\n",
"│ │ 512) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_9 (Conv2D) │ (None, None, None, │ 2,359,808 │\n",
"│ │ 512) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_10 (Conv2D) │ (None, None, None, │ 2,359,808 │\n",
"│ │ 512) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_11 (Conv2D) │ (None, None, None, │ 2,359,808 │\n",
"│ │ 512) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_12 (Conv2D) │ (None, None, None, │ 2,359,808 │\n",
"│ │ 512) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_13 (Conv2D) │ (None, None, None, │ 1,179,904 │\n",
"│ │ 256) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_14 (Conv2D) │ (None, None, None, │ 295,040 │\n",
"│ │ 128) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_15 (Conv2D) │ (None, None, None, 64) │ 73,792 │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_16 (Conv2D) │ (None, None, None, 1) │ 65 │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
"\n"
],
"text/plain": [
"┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
"┃\u001b[1m \u001b[0m\u001b[1mLayer (type) \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
"┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
"│ conv2d (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m1,792\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_1 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m36,928\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ max_pooling2d (\u001b[38;5;33mMaxPooling2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m0\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_2 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m73,856\u001b[0m │\n",
"│ │ \u001b[38;5;34m128\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_3 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m147,584\u001b[0m │\n",
"│ │ \u001b[38;5;34m128\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ max_pooling2d_1 (\u001b[38;5;33mMaxPooling2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m0\u001b[0m │\n",
"│ │ \u001b[38;5;34m128\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_4 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m295,168\u001b[0m │\n",
"│ │ \u001b[38;5;34m256\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_5 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m590,080\u001b[0m │\n",
"│ │ \u001b[38;5;34m256\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_6 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m590,080\u001b[0m │\n",
"│ │ \u001b[38;5;34m256\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ max_pooling2d_2 (\u001b[38;5;33mMaxPooling2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m0\u001b[0m │\n",
"│ │ \u001b[38;5;34m256\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_7 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m1,180,160\u001b[0m │\n",
"│ │ \u001b[38;5;34m512\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_8 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m2,359,808\u001b[0m │\n",
"│ │ \u001b[38;5;34m512\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_9 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m2,359,808\u001b[0m │\n",
"│ │ \u001b[38;5;34m512\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_10 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m2,359,808\u001b[0m │\n",
"│ │ \u001b[38;5;34m512\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_11 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m2,359,808\u001b[0m │\n",
"│ │ \u001b[38;5;34m512\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_12 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m2,359,808\u001b[0m │\n",
"│ │ \u001b[38;5;34m512\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_13 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m1,179,904\u001b[0m │\n",
"│ │ \u001b[38;5;34m256\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_14 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, │ \u001b[38;5;34m295,040\u001b[0m │\n",
"│ │ \u001b[38;5;34m128\u001b[0m) │ │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_15 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m64\u001b[0m) │ \u001b[38;5;34m73,792\u001b[0m │\n",
"├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
"│ conv2d_16 (\u001b[38;5;33mConv2D\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m) │ \u001b[38;5;34m65\u001b[0m │\n",
"└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Total params: 16,263,489 (62.04 MB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Total params: \u001b[0m\u001b[38;5;34m16,263,489\u001b[0m (62.04 MB)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Trainable params: 16,263,489 (62.04 MB)\n",
"\n"
],
"text/plain": [
"\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m16,263,489\u001b[0m (62.04 MB)\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
" Non-trainable params: 0 (0.00 B)\n",
"\n"
],
"text/plain": [
"\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"model.summary()"
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {},
"outputs": [],
"source": [
"train_gen = image_generator(img_paths,1)"
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Checking: data\\part_A_final/train_data\\ground\\IMG_1.h5\n",
"Exists: True\n"
]
}
],
"source": [
"# Test if density maps exist\n",
"test_path = img_paths[0].replace('.jpg', '.h5').replace('images', 'ground')\n",
"print(f\"Checking: {test_path}\")\n",
"print(f\"Exists: {os.path.exists(test_path)}\")\n"
]
},
{
"cell_type": "code",
"execution_count": 15,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Starting training...\n",
"Epoch 1/5\n",
"\u001b[1m20/20\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2057s\u001b[0m 104s/step - loss: 0.0270 - mse: 0.0270\n",
"Epoch 2/5\n",
"\u001b[1m20/20\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1935s\u001b[0m 96s/step - loss: 0.0596 - mse: 0.0596\n",
"Epoch 3/5\n",
"\u001b[1m20/20\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2106s\u001b[0m 103s/step - loss: 0.0830 - mse: 0.0830\n",
"Epoch 4/5\n",
"\u001b[1m20/20\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3105s\u001b[0m 158s/step - loss: 0.0255 - mse: 0.0255\n",
"Epoch 5/5\n",
"\u001b[1m20/20\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1841s\u001b[0m 92s/step - loss: 0.0355 - mse: 0.0355\n"
]
},
{
"data": {
"text/plain": [
""
]
},
"execution_count": 15,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"\n",
"# Train model (Fixed: use fit instead of fit_generator)\n",
"import numpy as np\n",
"print(\"Starting training...\")\n",
"model.fit(train_gen, epochs=5, steps_per_epoch=20, verbose=1)"
]
},
{
"cell_type": "code",
"execution_count": 20,
"metadata": {},
"outputs": [],
"source": [
"os.makedirs('weights', exist_ok=True)\n",
"os.makedirs('models', exist_ok=True)\n",
"save_mod(model, \"weights/model_A.weights.h5\", \"models/Model.json\")\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "base",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.13.5"
}
},
"nbformat": 4,
"nbformat_minor": 2
}