model_presentation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "initial_id",
   "metadata": {
    "collapsed": true,
    "ExecuteTime": {
     "end_time": "2023-11-07T19:04:02.205960800Z",
     "start_time": "2023-11-07T19:03:56.568333100Z"
    }
   },
   "outputs": [],
   "source": [
    "from model_cnn import MyNeuralNetwork\n",
    "import pickle\n",
    "filename = 'models/best_model.sav'\n",
    "model = pickle.load(open(filename, 'rb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "source": [
    "Poddamy krótkiej analizie model osiągający najlepsze wyniki, wytrenowany w zewnętrznym środowisku wykonawczym. Ma on cztery warstwy konwolucyjne w celu ekstrakcji cech obrazów i trzy warstwy liniowe zajmujące się klasyfikacją. "
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "19c0aa80e72b9caf"
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Struktura modelu: MyNeuralNetwork(\n",
      "  (conv_layers): Sequential(\n",
      "    (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (1): ReLU()\n",
      "    (2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
      "    (3): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (4): ReLU()\n",
      "    (5): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (6): ReLU()\n",
      "    (7): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (8): ReLU()\n",
      "    (9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
      "  )\n",
      "  (fc_layers): Sequential(\n",
      "    (0): Flatten(start_dim=1, end_dim=-1)\n",
      "    (1): Linear(in_features=3136, out_features=512, bias=True)\n",
      "    (2): ReLU()\n",
      "    (3): Linear(in_features=512, out_features=512, bias=True)\n",
      "    (4): ReLU()\n",
      "    (5): Linear(in_features=512, out_features=26, bias=True)\n",
      "  )\n",
      ")\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(f\"Struktura modelu: {model}\\n\\n\")"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2023-11-07T19:04:17.275028400Z",
     "start_time": "2023-11-07T19:04:17.258973500Z"
    }
   },
   "id": "c92529d169d5b88a"
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "outputs": [],
   "source": [
    "from data_functions import get_loaders_weights_and_occurrences, load_data\n",
    "_, val_loader, _, occurrences = get_loaders_weights_and_occurrences(32)\n",
    "_, _, X_val, y_val, _ = load_data()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2023-11-07T19:03:51.540781100Z",
     "start_time": "2023-11-07T19:03:43.885408400Z"
    }
   },
   "id": "36d3e41b13e9ec2b"
  },
  {
   "cell_type": "markdown",
   "source": [
    "Przeanalizujmy osiągi modelu na zbiorze walidacyjnym"
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "df14ad8dd2d0d154"
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 score for class A (occ. = 5126): 0.99\n",
      "F1 score for class B (occ. = 3102): 0.98\n",
      "F1 score for class C (occ. = 8075): 0.99\n",
      "F1 score for class D (occ. = 3650): 0.93\n",
      "F1 score for class E (occ. = 3947): 0.99\n",
      "F1 score for class F (occ. = 7345): 0.98\n",
      "F1 score for class G (occ. = 2014): 0.96\n",
      "F1 score for class H (occ. = 2522): 0.97\n",
      "F1 score for class I (occ. = 9557): 0.99\n",
      "F1 score for class J (occ. = 3010): 0.96\n",
      "F1 score for class K (occ. = 1974): 0.97\n",
      "F1 score for class L (occ. = 4061): 0.97\n",
      "F1 score for class M (occ. = 7201): 0.99\n",
      "F1 score for class N (occ. = 6590): 0.98\n",
      "F1 score for class O (occ. = 19986): 0.98\n",
      "F1 score for class P (occ. = 6678): 0.99\n",
      "F1 score for class Q (occ. = 2084): 0.96\n",
      "F1 score for class R (occ. = 4058): 0.98\n",
      "F1 score for class S (occ. = 16611): 0.99\n",
      "F1 score for class T (occ. = 7856): 0.99\n",
      "F1 score for class U (occ. = 10081): 0.97\n",
      "F1 score for class V (occ. = 3710): 0.92\n",
      "F1 score for class W (occ. = 3756): 0.97\n",
      "F1 score for class X (occ. = 2217): 0.97\n",
      "F1 score for class Y (occ. = 3794): 0.97\n",
      "F1 score for class Z (occ. = 2161): 0.98\n",
      "Accuracy on the test set: 97.93%\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import f1_score\n",
    "import torch\n",
    "\n",
    "model.eval()\n",
    "correct = 0\n",
    "total = 0\n",
    "true_labels = []\n",
    "predicted_labels = []\n",
    "\n",
    "with torch.no_grad():\n",
    "    for batch_data, labels in val_loader:\n",
    "        outputs = model(batch_data)\n",
    "        _, predicted = torch.max(outputs.data, 1)\n",
    "        total += labels.size(0)\n",
    "        correct += (predicted == labels).sum().item()\n",
    "\n",
    "        true_labels.extend(labels.tolist())\n",
    "        predicted_labels.extend(predicted.tolist())\n",
    "\n",
    "f1_scores = f1_score(true_labels, predicted_labels, average=None)\n",
    "from data_functions import class_mapping\n",
    "for class_idx, f1 in enumerate(f1_scores):\n",
    "    print(f'F1 score for class {class_mapping[class_idx]} (occ. = {occurrences[class_mapping[class_idx]]}): {f1:.2f}')\n",
    "\n",
    "print(f'Accuracy on the test set: {100 * correct / total:.2f}%')"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2023-11-07T12:44:12.231730500Z",
     "start_time": "2023-11-07T12:43:58.541947600Z"
    }
   },
   "id": "6a78ac9b84275473"
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wartość współczynnika korelacji (Pearson): 0.42064871522946196\n"
     ]
    },
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "corr_arr = np.array(\n",
    "    [[tensor.item() for tensor in occurrences.values()], [tensor.item() for tensor in f1_scores]]).T\n",
    "print(f'Wartość współczynnika korelacji (Pearson): {np.corrcoef(corr_arr[:, 0], corr_arr[:, 1])[0, 1]}')\n",
    "z = np.polyfit(list(occurrences.values()),f1_scores,1)\n",
    "p = np.poly1d(z)\n",
    "plt.plot(occurrences.values(),p(list(occurrences.values())), color='orange', linewidth=2)\n",
    "plt.scatter(occurrences.values(), f1_scores)\n",
    "plt.title(f\"Zależność osiągów od reprezentacji klasy w zbiorze treningowym\")\n",
    "plt.xlabel(\"Wystąpienia\")\n",
    "plt.ylabel(\"Wartość metryki f1\")\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2023-11-07T12:26:07.829214600Z",
     "start_time": "2023-11-07T12:26:07.657296600Z"
    }
   },
   "id": "f273291af4b235d6"
  },
  {
   "cell_type": "markdown",
   "source": [
    "Widzimy, że rozpoznawanie *D* i *V* sprawia modelowi szczególnie dużo trudności. Reszta wyników to ok. 98%. "
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "6715168bb9487e45"
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}