Visual Programming Project.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install opencv-python\n",
    "!pip install mediapipe\n",
    "!pip install scikit-learn"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Importing libraries and defining dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# For capturing hand coordinates\n",
    "import cv2\n",
    "import mediapipe as mp\n",
    "\n",
    "# For processing data\n",
    "import pandas as pd\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>class</th>\n",
       "      <th>x1</th>\n",
       "      <th>y1</th>\n",
       "      <th>x2</th>\n",
       "      <th>y2</th>\n",
       "      <th>x3</th>\n",
       "      <th>y3</th>\n",
       "      <th>x4</th>\n",
       "      <th>y4</th>\n",
       "      <th>x5</th>\n",
       "      <th>...</th>\n",
       "      <th>x17</th>\n",
       "      <th>y17</th>\n",
       "      <th>x18</th>\n",
       "      <th>y18</th>\n",
       "      <th>x19</th>\n",
       "      <th>y19</th>\n",
       "      <th>x20</th>\n",
       "      <th>y20</th>\n",
       "      <th>x21</th>\n",
       "      <th>y21</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>A</td>\n",
       "      <td>0.295849</td>\n",
       "      <td>0.679613</td>\n",
       "      <td>0.209038</td>\n",
       "      <td>0.612926</td>\n",
       "      <td>0.147883</td>\n",
       "      <td>0.522415</td>\n",
       "      <td>0.125748</td>\n",
       "      <td>0.439133</td>\n",
       "      <td>0.130589</td>\n",
       "      <td>...</td>\n",
       "      <td>0.295946</td>\n",
       "      <td>0.572083</td>\n",
       "      <td>0.362726</td>\n",
       "      <td>0.507097</td>\n",
       "      <td>0.362393</td>\n",
       "      <td>0.452241</td>\n",
       "      <td>0.351849</td>\n",
       "      <td>0.504388</td>\n",
       "      <td>0.340516</td>\n",
       "      <td>0.553928</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>A</td>\n",
       "      <td>0.277914</td>\n",
       "      <td>0.728127</td>\n",
       "      <td>0.168684</td>\n",
       "      <td>0.634556</td>\n",
       "      <td>0.107125</td>\n",
       "      <td>0.489586</td>\n",
       "      <td>0.104561</td>\n",
       "      <td>0.368203</td>\n",
       "      <td>0.126103</td>\n",
       "      <td>...</td>\n",
       "      <td>0.284394</td>\n",
       "      <td>0.614332</td>\n",
       "      <td>0.458914</td>\n",
       "      <td>0.482608</td>\n",
       "      <td>0.402338</td>\n",
       "      <td>0.455794</td>\n",
       "      <td>0.359927</td>\n",
       "      <td>0.538472</td>\n",
       "      <td>0.352813</td>\n",
       "      <td>0.605610</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>A</td>\n",
       "      <td>0.273304</td>\n",
       "      <td>0.750424</td>\n",
       "      <td>0.153818</td>\n",
       "      <td>0.649119</td>\n",
       "      <td>0.087932</td>\n",
       "      <td>0.492969</td>\n",
       "      <td>0.090790</td>\n",
       "      <td>0.370735</td>\n",
       "      <td>0.114260</td>\n",
       "      <td>...</td>\n",
       "      <td>0.292044</td>\n",
       "      <td>0.643904</td>\n",
       "      <td>0.450758</td>\n",
       "      <td>0.495144</td>\n",
       "      <td>0.392497</td>\n",
       "      <td>0.475129</td>\n",
       "      <td>0.353110</td>\n",
       "      <td>0.558542</td>\n",
       "      <td>0.348972</td>\n",
       "      <td>0.624010</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>A</td>\n",
       "      <td>0.260779</td>\n",
       "      <td>0.758578</td>\n",
       "      <td>0.142449</td>\n",
       "      <td>0.658145</td>\n",
       "      <td>0.081542</td>\n",
       "      <td>0.496756</td>\n",
       "      <td>0.090102</td>\n",
       "      <td>0.373770</td>\n",
       "      <td>0.115699</td>\n",
       "      <td>...</td>\n",
       "      <td>0.282037</td>\n",
       "      <td>0.659344</td>\n",
       "      <td>0.449754</td>\n",
       "      <td>0.502502</td>\n",
       "      <td>0.380867</td>\n",
       "      <td>0.484203</td>\n",
       "      <td>0.343782</td>\n",
       "      <td>0.567716</td>\n",
       "      <td>0.341316</td>\n",
       "      <td>0.632237</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>A</td>\n",
       "      <td>0.264733</td>\n",
       "      <td>0.761643</td>\n",
       "      <td>0.144033</td>\n",
       "      <td>0.670212</td>\n",
       "      <td>0.077731</td>\n",
       "      <td>0.505519</td>\n",
       "      <td>0.080886</td>\n",
       "      <td>0.379795</td>\n",
       "      <td>0.101522</td>\n",
       "      <td>...</td>\n",
       "      <td>0.289165</td>\n",
       "      <td>0.674617</td>\n",
       "      <td>0.432672</td>\n",
       "      <td>0.504809</td>\n",
       "      <td>0.386974</td>\n",
       "      <td>0.488224</td>\n",
       "      <td>0.355972</td>\n",
       "      <td>0.573528</td>\n",
       "      <td>0.350644</td>\n",
       "      <td>0.638745</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 43 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "  class        x1        y1        x2        y2        x3        y3        x4  \\\n",
       "0     A  0.295849  0.679613  0.209038  0.612926  0.147883  0.522415  0.125748   \n",
       "1     A  0.277914  0.728127  0.168684  0.634556  0.107125  0.489586  0.104561   \n",
       "2     A  0.273304  0.750424  0.153818  0.649119  0.087932  0.492969  0.090790   \n",
       "3     A  0.260779  0.758578  0.142449  0.658145  0.081542  0.496756  0.090102   \n",
       "4     A  0.264733  0.761643  0.144033  0.670212  0.077731  0.505519  0.080886   \n",
       "\n",
       "         y4        x5  ...       x17       y17       x18       y18       x19  \\\n",
       "0  0.439133  0.130589  ...  0.295946  0.572083  0.362726  0.507097  0.362393   \n",
       "1  0.368203  0.126103  ...  0.284394  0.614332  0.458914  0.482608  0.402338   \n",
       "2  0.370735  0.114260  ...  0.292044  0.643904  0.450758  0.495144  0.392497   \n",
       "3  0.373770  0.115699  ...  0.282037  0.659344  0.449754  0.502502  0.380867   \n",
       "4  0.379795  0.101522  ...  0.289165  0.674617  0.432672  0.504809  0.386974   \n",
       "\n",
       "        y19       x20       y20       x21       y21  \n",
       "0  0.452241  0.351849  0.504388  0.340516  0.553928  \n",
       "1  0.455794  0.359927  0.538472  0.352813  0.605610  \n",
       "2  0.475129  0.353110  0.558542  0.348972  0.624010  \n",
       "3  0.484203  0.343782  0.567716  0.341316  0.632237  \n",
       "4  0.488224  0.355972  0.573528  0.350644  0.638745  \n",
       "\n",
       "[5 rows x 43 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset = pd.read_csv('./Dataset/hand_dataset_1000_24.csv')\n",
    "\n",
    "# Show dataset first five data\n",
    "dataset.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "K    1000\n",
       "D    1000\n",
       "X    1000\n",
       "S    1000\n",
       "G    1000\n",
       "U    1000\n",
       "R    1000\n",
       "A    1000\n",
       "Q    1000\n",
       "N    1000\n",
       "V    1000\n",
       "P    1000\n",
       "E    1000\n",
       "W    1000\n",
       "H    1000\n",
       "T    1000\n",
       "M    1000\n",
       "I    1000\n",
       "L    1000\n",
       "O    1000\n",
       "F    1000\n",
       "B    1000\n",
       "Y    1000\n",
       "C    1000\n",
       "Name: class, dtype: int64"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Show dataset overview, should return 1000 for each alphabet (excluding y and z)\n",
    "dataset['class'].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Creating Train and Test Data\n",
    "- We use *train_test_split* since we don't really have test dataset.\n",
    "- Normalizing dataset can be ignored, since we predict our data directly using raw handlandmark. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Defining X and Y from dataset for training and testing\n",
    "\n",
    "X = dataset.iloc[:, 1:].values\n",
    "Y = dataset.iloc[:, 0].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# We will take 33% from 1000 for our test data.\n",
    "# Recommended value 80:20, 67:33, 50:50\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, Y, test_size=0.33)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Normalize / Standarize dataset\n",
    "\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "scaler = StandardScaler().fit(X_train)\n",
    "\n",
    "X_train = scaler.transform(X_train)\n",
    "X_test = scaler.transform(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3. Creating classifier model for our alphabet recognition.\n",
    "- *n_neighbors* can be adjusted as we provide graph for mean errors for each *n_neighbors*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "KNeighborsClassifier(n_neighbors=3)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "classifier = KNeighborsClassifier(n_neighbors=3)\n",
    "classifier.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 4. Calculate model accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = classifier.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           A       0.93      0.99      0.96       297\n",
      "           B       0.96      0.99      0.97       325\n",
      "           C       0.97      0.99      0.98       335\n",
      "           D       0.98      0.97      0.97       339\n",
      "           E       0.93      0.97      0.95       320\n",
      "           F       0.98      0.98      0.98       316\n",
      "           G       0.98      0.99      0.98       361\n",
      "           H       0.98      0.99      0.99       328\n",
      "           I       0.97      0.97      0.97       343\n",
      "           K       0.94      0.98      0.96       332\n",
      "           L       0.98      0.98      0.98       330\n",
      "           M       0.85      0.84      0.84       311\n",
      "           N       0.85      0.84      0.85       332\n",
      "           O       0.97      0.93      0.95       339\n",
      "           P       0.96      0.94      0.95       327\n",
      "           Q       0.94      0.96      0.95       322\n",
      "           R       0.90      0.91      0.91       321\n",
      "           S       0.93      0.92      0.93       292\n",
      "           T       0.97      0.94      0.96       342\n",
      "           U       0.84      0.86      0.85       331\n",
      "           V       0.92      0.84      0.88       339\n",
      "           W       0.99      0.97      0.98       349\n",
      "           X       0.94      0.92      0.93       343\n",
      "           Y       0.98      0.94      0.96       346\n",
      "\n",
      "    accuracy                           0.94      7920\n",
      "   macro avg       0.94      0.94      0.94      7920\n",
      "weighted avg       0.94      0.94      0.94      7920\n",
      "\n",
      "0.9440656565656566\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report, accuracy_score\n",
    "print(classification_report(y_test, y_pred))\n",
    "print(accuracy_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 5. Show graph for adjusting number of *n_neighbors*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "error = []\n",
    "\n",
    "# Calculating error for K values between 1 and 40\n",
    "for i in range(1, 40):\n",
    "    knn = KNeighborsClassifier(n_neighbors=i)\n",
    "    knn.fit(X_train, y_train)\n",
    "    pred_i = knn.predict(X_test)\n",
    "    error.append(np.mean(pred_i != y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'Mean Error')"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.figure(figsize=(12, 6))\n",
    "plt.plot(range(1, 40), error, color='red', linestyle='dashed', marker='o',\n",
    "         markerfacecolor='blue', markersize=10)\n",
    "plt.title('Error Rate K Value')\n",
    "plt.xlabel('K Value')\n",
    "plt.ylabel('Mean Error')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 6. Intialize Mediapipe Hands for alphabet recognition."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initialize mediapipe hand\n",
    "\n",
    "mp_drawing = mp.solutions.drawing_utils\n",
    "mp_hands = mp.solutions.hands"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initialize mediapipe hand capture webcam\n",
    "\n",
    "cap = cv2.VideoCapture(0)\n",
    "with mp_hands.Hands(\n",
    "    max_num_hands = 1,\n",
    "    min_detection_confidence=0.5,\n",
    "    min_tracking_confidence=0.5) as hands:\n",
    "    while cap.isOpened():\n",
    "        success, image = cap.read()\n",
    "\n",
    "        if not success:\n",
    "            print(\"Ignoring empty camera frame.\")\n",
    "            # If loading a video, use 'break' instead of 'continue'.\n",
    "            continue\n",
    "\n",
    "        # Flip the image horizontally for a later selfie-view display, and convert\n",
    "        # the BGR image to RGB.\n",
    "        image = cv2.cvtColor(cv2.flip(image, 1), cv2.COLOR_BGR2RGB)\n",
    "\n",
    "        # To improve performance, optionally mark the image as not writeable to\n",
    "        # pass by reference.\n",
    "        image.flags.writeable = False\n",
    "        results = hands.process(image)\n",
    "\n",
    "        # Draw the hand annotations on the image.\n",
    "        image.flags.writeable = True\n",
    "        image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)\n",
    "        if results.multi_hand_landmarks:\n",
    "            for hand_landmarks in results.multi_hand_landmarks:\n",
    "                coords = hand_landmarks.landmark\n",
    "                mp_drawing.draw_landmarks(image, hand_landmarks, mp_hands.HAND_CONNECTIONS)\n",
    "                coords = list(np.array([[landmark.x, landmark.y] for landmark in coords]).flatten())\n",
    "                coords = scaler.transform([coords])\n",
    "                \n",
    "                # Alternative for dataset using z coordinates.\n",
    "                # Z coordinates is not recommended, since you need to adjust your distance from camera.\n",
    "#                 coords = list(np.array([[landmark.x, landmark.y, landmark.z] for landmark in coords]).flatten())\n",
    "                \n",
    "                predicted = classifier.predict(coords)\n",
    "\n",
    "            # Get status box\n",
    "            cv2.rectangle(image, (0,0), (100, 60), (245, 90, 16), -1)\n",
    "\n",
    "            # Display Class\n",
    "            cv2.putText(image, 'CLASS'\n",
    "                        , (20,15), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0), 1, cv2.LINE_AA)\n",
    "            cv2.putText(image, str(predicted[0])\n",
    "                        , (20,45), cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255), 2, cv2.LINE_AA)\n",
    "\n",
    "        cv2.imshow('MediaPipe Hands', image)\n",
    "\n",
    "        # Press esc to close webcam\n",
    "        if cv2.waitKey(5) & 0xFF == 27:\n",
    "            break\n",
    "cap.release()\n",
    "cv2.destroyAllWindows()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}