model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "is_executing": true
   },
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras import layers\n",
    "from collections import Counter\n",
    "\n",
    "from tensorflow.keras.preprocessing.image import load_img, img_to_array\n",
    "from sklearn.model_selection import train_test_split\n",
    "from concurrent.futures import ProcessPoolExecutor\n",
    "\n",
    "import random"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "TARGET_SIZE = (64, 64)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def parse_filename(filename):\n",
    "    components = filename.split(\"_\")\n",
    "    offset = 0\n",
    "    \n",
    "    if \"-\" in components[1]:\n",
    "        offset = 1\n",
    "    \n",
    "    if len(components) < 6 + offset:\n",
    "        return False, []\n",
    "   \n",
    "    x = int(components[1 + offset])\n",
    "    y = int(components[2 + offset])\n",
    "    z = int(components[3 + offset])\n",
    "    r = int(components[4 + offset])\n",
    "    is_flying = int(components[5 + offset].split(\".\")[0])\n",
    "\n",
    "    if x == 0 and y == 0 and z == 0 and r == 0:\n",
    "        return False, []\n",
    "\n",
    "    return True, [x, y, z, r, is_flying]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "image_data = []\n",
    "label_data = []\n",
    "\n",
    "for date_folder in os.listdir(\"data\"):\n",
    "    date_folder_path = os.path.join(\"data\", date_folder)\n",
    "\n",
    "    if os.path.isdir(date_folder_path):\n",
    "        for filename in os.listdir(date_folder_path):\n",
    "            if filename.endswith(\".png\"):\n",
    "                image_path = os.path.join(date_folder_path, filename)\n",
    "                valid, components = parse_filename(filename)\n",
    "                \n",
    "                if valid:\n",
    "                    image_data.append(image_path)\n",
    "                    label_data.append(components)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "count = Counter([tuple(x) for x in label_data])\n",
    "count = count.most_common()\n",
    "\n",
    "count = [x for x in count if x[1] >= 700]\n",
    "\n",
    "f_image_data = [image_data[i] for i in range(len(label_data)) if tuple(label_data[i]) in [x[0] for x in count]]\n",
    "f_label_data = [label_data[i] for i in range(len(label_data)) if tuple(label_data[i]) in [x[0] for x in count]]\n",
    "\n",
    "f_label_data = [\"_\".join([str(x) for x in y]) for y in f_label_data]\n",
    "\n",
    "tokenizer = {x: i for i, x in enumerate(set(f_label_data))}\n",
    "detokenizer = {i: x for i, x in enumerate(set(f_label_data))}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# make a histogram plot of the data\n",
    "plt.hist(f_label_data, bins=len(tokenizer))\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def preprocess_image(image_path):\n",
    "    image = load_img(image_path, target_size=TARGET_SIZE) \n",
    "    image = img_to_array(image) / 255.0\n",
    "\n",
    "    return image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def preprocess_and_cache_images(image_paths, cache_folder, max_workers):\n",
    "    os.makedirs(cache_folder, exist_ok=True)\n",
    "    cached_images = []\n",
    "\n",
    "    with ProcessPoolExecutor(max_workers=max_workers) as executor:\n",
    "        for image_path in image_paths:\n",
    "            cached_path = os.path.join(cache_folder, os.path.basename(image_path) + \".npy\")\n",
    "            if os.path.exists(cached_path):\n",
    "                cached_images.append(np.load(cached_path))\n",
    "            else:\n",
    "                image = preprocess_image(image_path)\n",
    "                np.save(cached_path, image)\n",
    "                cached_images.append(cached_path)\n",
    "\n",
    "    return cached_images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "f_image_data = preprocess_and_cache_images(f_image_data, \"cache\", max_workers=os.cpu_count())\n",
    "one_hot_labels = tf.keras.utils.to_categorical([tokenizer[x] for x in f_label_data])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "11849 11849 8082 8082\n"
     ]
    }
   ],
   "source": [
    "print(len(label_data), len(image_data), len(f_label_data), len(f_image_data))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6465, 64, 64, 3) (1617, 64, 64, 3) (6465, 6) (1617, 6)\n"
     ]
    }
   ],
   "source": [
    "image_train, image_val, label_train, label_val = train_test_split(\n",
    "    f_image_data, one_hot_labels, test_size=0.2, random_state=42)\n",
    "\n",
    "image_train = np.array(image_train)\n",
    "image_val = np.array(image_val)\n",
    "label_train = np.array(label_train)\n",
    "label_val = np.array(label_val)\n",
    "\n",
    "print(image_train.shape, image_val.shape, label_train.shape, label_val.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0_2_0_0_1\n"
     ]
    }
   ],
   "source": [
    "# display a random picture from the dataset + label\n",
    "random_index = random.randint(0, len(image_train))\n",
    "plt.imshow(image_train[random_index])\n",
    "plt.show()\n",
    "print(detokenizer[np.argmax(label_train[random_index])])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(TARGET_SIZE[0], TARGET_SIZE[1], 3)))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Dropout(0.4))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.Dropout(0.4))\n",
    "model.add(layers.Conv2D(128, (3, 3), activation='relu'))\n",
    "model.add(layers.Dropout(0.4))\n",
    "\n",
    "model.add(layers.Flatten())\n",
    "\n",
    "model.add(layers.Dense(128, activation='relu'))\n",
    "model.add(layers.Dropout(0.4))\n",
    "model.add(layers.Dense(len(tokenizer), activation='softmax'))\n",
    "\n",
    "# model.add(layers.Conv2D(128, (3, 3), , activation=\"relu\", padding=\"same\"))\n",
    "# model.add(layers.MaxPooling2D(pool_size=(3, 3)))\n",
    "# model.add(layers.Conv2D(64, (3, 3), activation=\"relu\", padding=\"same\"))\n",
    "# model.add(layers.MaxPooling2D(pool_size=(3, 3)))\n",
    "# model.add(layers.Conv2D(32, (3, 3), activation=\"relu\", padding=\"same\"))\n",
    "# model.add(layers.MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "# model.add(layers.Flatten())\n",
    "\n",
    "# model.add(layers.Dense(64, activation=\"relu\"))\n",
    "# model.add(layers.Dense(len(tokenizer), activation=\"softmax\")) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(\n",
    "    optimizer=\"adam\",\n",
    "    loss=\"categorical_crossentropy\",\n",
    "    metrics=[\"accuracy\"]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "class DataGenerator(tf.keras.utils.Sequence):\n",
    "    def __init__(self, image_paths, labels, batch_size, shuffle=True):\n",
    "        self.image_paths = image_paths\n",
    "        self.labels = labels\n",
    "        self.batch_size = batch_size\n",
    "        self.shuffle = shuffle\n",
    "        self.on_epoch_end()\n",
    "\n",
    "    def __len__(self):\n",
    "        return int(np.floor(len(self.image_paths) / self.batch_size))\n",
    "\n",
    "    def __getitem__(self, index):\n",
    "        indexes = self.indexes[index * self.batch_size:(index + 1) * self.batch_size]\n",
    "        image_paths_temp = [self.image_paths[k] for k in indexes]\n",
    "        labels_temp = [self.labels[k] for k in indexes]\n",
    "\n",
    "        images = []\n",
    "        for image_path in image_paths_temp:\n",
    "            if isinstance(image_path, str):\n",
    "                images.append(np.load(image_path))\n",
    "            else:\n",
    "                images.append(image_path)\n",
    "\n",
    "        return np.array(images), np.array(labels_temp)\n",
    "\n",
    "    def on_epoch_end(self):\n",
    "        self.indexes = np.arange(len(self.image_paths))\n",
    "        if self.shuffle == True:\n",
    "            np.random.shuffle(self.indexes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/8\n",
      "50/50 [==============================] - 1s 27ms/step - loss: 0.2038 - accuracy: 0.9162 - val_loss: 0.9334 - val_accuracy: 0.8535\n",
      "Epoch 2/8\n",
      "50/50 [==============================] - 1s 24ms/step - loss: 0.2130 - accuracy: 0.9134 - val_loss: 0.9537 - val_accuracy: 0.8509\n",
      "Epoch 3/8\n",
      "50/50 [==============================] - 1s 24ms/step - loss: 0.1960 - accuracy: 0.9230 - val_loss: 0.9260 - val_accuracy: 0.8503\n",
      "Epoch 4/8\n",
      "50/50 [==============================] - 1s 24ms/step - loss: 0.1974 - accuracy: 0.9239 - val_loss: 0.9231 - val_accuracy: 0.8522\n",
      "Epoch 5/8\n",
      "50/50 [==============================] - 1s 24ms/step - loss: 0.1943 - accuracy: 0.9203 - val_loss: 0.8966 - val_accuracy: 0.8529\n",
      "Epoch 6/8\n",
      "50/50 [==============================] - 1s 24ms/step - loss: 0.2007 - accuracy: 0.9169 - val_loss: 1.0299 - val_accuracy: 0.8438\n",
      "Epoch 7/8\n",
      "50/50 [==============================] - 1s 24ms/step - loss: 0.1932 - accuracy: 0.9216 - val_loss: 0.9831 - val_accuracy: 0.8405\n",
      "Epoch 8/8\n",
      "50/50 [==============================] - 1s 24ms/step - loss: 0.1898 - accuracy: 0.9264 - val_loss: 1.0770 - val_accuracy: 0.8431\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x2382fbadd20>"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_generator = DataGenerator(image_train, label_train, batch_size=128)\n",
    "val_generator = DataGenerator(image_val, label_val, batch_size=128)\n",
    "\n",
    "model.fit(\n",
    "    train_generator,\n",
    "    epochs=8,\n",
    "    validation_data=val_generator\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "directory = \"models/COPILOTVA85_11-1-23\"\n",
    "os.makedirs(directory, exist_ok=False)\n",
    "\n",
    "model.save(f\"{directory}/model.h5\")\n",
    "np.save(f\"{directory}/tokenizer.npy\", tokenizer)\n",
    "np.save(f\"{directory}/detokenizer.npy\", detokenizer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# def predict(filename, from_folder=False):\n",
    "#     image = None\n",
    "\n",
    "#     if from_folder:\n",
    "#         image = preprocess_image(filename)\n",
    "\n",
    "#     image = np.expand_dims(image or filename, axis=0)\n",
    "\n",
    "#     if not from_folder:\n",
    "#         plt.imshow(image[0])\n",
    "\n",
    "#     predictions = model.predict(image)[0]\n",
    "#     predictions = np.argmax(predictions)\n",
    "\n",
    "#     result = detokenizer[predictions]\n",
    "\n",
    "#     return result\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ran_index = random.randint(0, len(image_val))\n",
    "# predictions = predict(image_val[ran_index], from_folder=False)\n",
    "# label = label_val[ran_index]\n",
    "\n",
    "# print(f\"predictions | {predictions}\")\n",
    "# print(f\"label       | {detokenizer[np.argmax(label)]}\")\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.10.10"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}