examples/anime_example.ipynb

{
 "cells": [
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "## This example show the ability of CNN with encoder-decoder architecture to approximate complex media data with periodic component.  ",
   "id": "89c6d2f14020350f"
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "#### As data source gif-file with Menhera manga character was used. Gif-file contains 10 frames (3 channel RGB images). Resolution of images was reduced to 45x45 pixels to light the task.",
   "id": "b816f50889ff4fb6"
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "![SegmentLocal](tools/media/anime_10f_fullsize.gif \"segment\")",
   "id": "7778517c8d6a362e"
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "#### Function to prepare dataset from gif-file to numpy matrices  ",
   "id": "cff823cde99499ca"
  },
  {
   "cell_type": "code",
   "id": "e6a3d5d05db5d1ab",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-10-14T13:47:26.773281Z",
     "start_time": "2024-10-14T13:47:26.371090Z"
    }
   },
   "source": [
    "import numpy as np\n",
    "from PIL import Image, ImageSequence\n",
    "\n",
    "def get_anime_timeseries(rgb=True):\n",
    "    with Image.open('tools/media/anime_10f.gif') as im:\n",
    "        array = []\n",
    "        for frame in ImageSequence.Iterator(im):\n",
    "            if rgb:\n",
    "                im_data = frame.copy().convert('RGB').getdata()\n",
    "                im_array = np.array(im_data).reshape(frame.size[1], frame.size[0], 3)\n",
    "            else:\n",
    "                im_data = frame.copy().convert('L').getdata()\n",
    "                im_array = np.array(im_data).reshape(frame.size[1], frame.size[0], 1)\n",
    "            array.append(im_array)\n",
    "        array = np.array(array)        \n",
    "        array = array/255\n",
    "    return array\n",
    "\n",
    "\n",
    "def get_cycled_data(cycles_num, is_rgb):\n",
    "    array = get_anime_timeseries(rgb=is_rgb)\n",
    "    arr = []\n",
    "    for i in range(cycles_num):\n",
    "        arr.append(array)\n",
    "    arr = np.array(arr)\n",
    "    arr = arr.reshape(arr.shape[0]*arr.shape[1], arr.shape[2], arr.shape[3], arr.shape[4])\n",
    "    return arr"
   ],
   "outputs": [],
   "execution_count": 1
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "### Train / test set initialization",
   "id": "358b31f37a88b22"
  },
  {
   "cell_type": "code",
   "id": "61165453374fda73",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-10-14T13:47:27.408795Z",
     "start_time": "2024-10-14T13:47:26.774283Z"
    }
   },
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "train = get_cycled_data(5, is_rgb=False)[:, :, :, 0]\n",
    "print(f'Train shape: {train.shape}')\n",
    "test = get_cycled_data(1, is_rgb=False)[:, :, :, 0]\n",
    "print(f'Test shape: {test.shape}')\n",
    "\n",
    "\n",
    "plt.imshow(train[3], cmap='Greys_r')\n",
    "plt.colorbar()\n",
    "plt.show()"
   ],
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train shape: (50, 45, 45)\n",
      "Test shape: (10, 45, 45)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "execution_count": 2
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "### Preparation datasets with moving window depend on prehistory and forecast horizon",
   "id": "364fb9afaecefeb1"
  },
  {
   "cell_type": "code",
   "id": "4fdf405ba3465604",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-10-14T13:47:37.510282Z",
     "start_time": "2024-10-14T13:47:34.330504Z"
    }
   },
   "source": [
    "from torchcnnbuilder.preprocess.time_series import multi_output_tensor, single_output_tensor\n",
    "\n",
    "train_dataset = multi_output_tensor(data=train,\n",
    "                                    pre_history_len=20,\n",
    "                                    forecast_len=10)\n",
    "test_dataset = single_output_tensor(data=test, \n",
    "                                   forecast_len=10)"
   ],
   "outputs": [],
   "execution_count": 3
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "### Simple model (with 5 layers of convolutions and 5 layers of transpose convolutions) initialization - 20 frames as prehistory, 10 frames as forecast horizon.",
   "id": "e94944db3d96a166"
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-10-14T13:51:57.747851Z",
     "start_time": "2024-10-14T13:51:57.547895Z"
    }
   },
   "cell_type": "code",
   "source": [
    "from torch import nn\n",
    "from torchcnnbuilder.models import ForecasterBase\n",
    "\n",
    "model = ForecasterBase(input_size=[45, 45],\n",
    "                       in_time_points=20,\n",
    "                       out_time_points=10,\n",
    "                       n_layers=5,\n",
    "                       finish_activation_function=nn.ReLU(inplace=True))\n",
    "device='cuda'\n",
    "model=model.to(device)"
   ],
   "id": "bcea5d367faf7a3e",
   "outputs": [],
   "execution_count": 4
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "### Train parameters set ",
   "id": "81a0c18bb9762c6a"
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-10-14T14:13:30.181383Z",
     "start_time": "2024-10-14T14:13:29.281678Z"
    }
   },
   "cell_type": "code",
   "source": [
    "from torch import optim\n",
    "from torch.utils.data import DataLoader\n",
    "\n",
    "epochs = 100000\n",
    "batch_size = 500\n",
    "\n",
    "dataloader = DataLoader(train_dataset, batch_size=batch_size, shuffle=False)\n",
    "optimizer = optim.Adam(model.parameters(), lr=1e-4)\n",
    "criterion = nn.L1Loss()\n",
    "losses = []\n",
    "epochs_list = []"
   ],
   "id": "5bb4b425f1a33233",
   "outputs": [],
   "execution_count": 5
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "### Training the model with progress save ",
   "id": "9328de589692b60d"
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-10-14T14:25:59.822849Z",
     "start_time": "2024-10-14T14:19:12.883444Z"
    }
   },
   "cell_type": "code",
   "source": [
    "from tqdm.notebook import tqdm\n",
    "import time\n",
    "import os\n",
    "\n",
    "# create folder for predicted images save\n",
    "root = os.path.abspath(\"\")\n",
    "opt_hist_path = f'{root}/anime_opt_hist_images'\n",
    "if not os.path.exists(opt_hist_path):\n",
    "    os.makedirs(opt_hist_path)\n",
    "\n",
    "start = time.time()\n",
    "for epoch in tqdm(range(epochs)):\n",
    "    loss = 0\n",
    "    \n",
    "    for train_features, test_features in dataloader:\n",
    "        train_features = train_features.to(device)\n",
    "        test_features = test_features.to(device)\n",
    "        \n",
    "        optimizer.zero_grad()\n",
    "        outputs = model(train_features)    \n",
    "        train_loss = criterion(outputs, test_features)        \n",
    "        train_loss.backward()\n",
    "        optimizer.step()        \n",
    "        loss += train_loss.item()\n",
    "    \n",
    "    loss = loss / len(dataloader)    \n",
    "    losses.append(loss)\n",
    "    epochs_list.append(epoch)\n",
    "    \n",
    "    # here we save optimization history as predicted images to visualize convergence process\n",
    "    if epoch % 10000 == 0 or epoch == 5 or epoch==50 or epoch==300 or epoch == 1000 or epoch == 5000:\n",
    "        for X, y in dataloader:\n",
    "            X = X.to(device)\n",
    "            prediction = model(X)\n",
    "            prediction = prediction.detach().cpu().numpy()[0]\n",
    "            real = y.numpy()[0]\n",
    "    \n",
    "            fig, (axs) = plt.subplots(2, 10, figsize=(10, 3))\n",
    "            for i in range(10):\n",
    "                axs[1, i].imshow(prediction[i], cmap='Greys_r', vmax=1, vmin=0)\n",
    "                axs[1, i].set_title(F'Frame {i}')\n",
    "                axs[0, i].imshow(real[i], cmap='Greys_r', vmax=1, vmin=0)\n",
    "                axs[0, i].set_title(F'Frame {i}')\n",
    "                axs[0, i].set_xticks([])\n",
    "                axs[1, i].set_xticks([])\n",
    "                axs[0, i].set_yticks([])\n",
    "                axs[1, i].set_yticks([])\n",
    "            plt.suptitle(f'Epoch={epoch}, loss={round(loss, 3)}')\n",
    "            plt.tight_layout()\n",
    "            plt.savefig(f'{opt_hist_path}/test_images_{epoch}.png')\n",
    "            plt.close()\n",
    "\n",
    "end = time.time()\n",
    "print(f'time spent: {end-start}')"
   ],
   "id": "1075054a695b6ee0",
   "outputs": [
    {
     "data": {
      "text/plain": [
       "  0%|          | 0/100000 [00:00<?, ?it/s]"
      ],
      "application/vnd.jupyter.widget-view+json": {
       "version_major": 2,
       "version_minor": 0,
       "model_id": "d2286f7f2e51401db16a1a50a0787ca9"
      }
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time spent: 406.92840790748596\n"
     ]
    }
   ],
   "execution_count": 10
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "Now in folder ***anime_opt_hist_images*** we have visualization of convergence process.  Saved png processed to gif: ",
   "id": "be7ac8e76e885ad1"
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "![SegmentLocal](tools/media/anime_convergence.gif \"segment\")",
   "id": "c81ce9fcba439e7"
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "### Lets visualize convergence plot",
   "id": "587e7a185a3cf2f8"
  },
  {
   "cell_type": "code",
   "id": "9f589401e37f5182",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-10-14T14:34:28.896386Z",
     "start_time": "2024-10-14T14:34:28.798344Z"
    }
   },
   "source": [
    "plt.plot(epochs_list, losses)\n",
    "plt.grid()\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('L1 Loss')\n",
    "plt.title('Loss history')\n",
    "plt.show()"
   ],
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "execution_count": 12
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "#### Conclusion",
   "id": "6c93dd802f1f3600"
  },
  {
   "cell_type": "code",
   "id": "8b2b11a97ed71d17",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-10-14T14:37:05.521583Z",
     "start_time": "2024-10-14T14:37:05.516581Z"
    }
   },
   "source": [
    "params = sum(p.numel() for p in model.parameters() if p.requires_grad)\n",
    "print(f'Total trainable parameters of the model: {params}')"
   ],
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total trainable parameters of the model: 72990\n"
     ]
    }
   ],
   "execution_count": 14
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "In this case media data can be determined as synthetic time series with periodic component. As we can see from convergence plot and prediction visualizations during training, such simple model (less than 100k parameters) has ability to correctly approximate complex synthetic data.   ",
   "id": "59d13d619d31c55e"
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}