Added DTW KNN notebooks

.gitignore

@@ -3,3 +3,16 @@
 /poses
 __pycache__/
 *.pyc
+.ipynb_checkpoints/analysis_shoulderpress-checkpoint.ipynb
+.ipynb_checkpoints/analysis_shoulderpress-checkpoint.ipynb
+.ipynb_checkpoints/DTW-FrontRaise-checkpoint.ipynb
+.ipynb_checkpoints/analysis_shoulderpress-checkpoint.ipynb
+.ipynb_checkpoints/analysis_shoulderpress-checkpoint.ipynb
+.ipynb_checkpoints/DTW-ShoulderPress-checkpoint.ipynb
+.ipynb_checkpoints/DTW-ShoulderShrug-checkpoint.ipynb
+.ipynb_checkpoints/ShoulderPress-checkpoint.ipynb
+.ipynb_checkpoints/ShoulderPressAnalysis-checkpoint.ipynb
+__init__.py
+.ipynb_checkpoints/analysis_shoulderpress-checkpoint.ipynb
+ml_notebooks/.ipynb_checkpoints/DTW-FrontRaise-checkpoint.ipynb
+ml_notebooks/.ipynb_checkpoints/DTW-ShoulderPress-Copy1-checkpoint.ipynb

DTW-FrontRaise.ipynb

@@ -0,0 +1,146 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "import math\n",
+    "import glob\n",
+    "import utils\n",
+    "\n",
+    "from parse import load_ps\n",
+    "from pprint import pprint\n",
+    "from scipy.signal import medfilt\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.neighbors import KNeighborsClassifier\n",
+    "from sklearn.neighbors import DistanceMetric\n",
+    "from sklearn.metrics import classification_report"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "files = utils.files_in_order('poses_compressed/frontraise')\n",
+    "X_train_names, X_test_names = train_test_split(files, test_size=0.4, random_state=42)\n",
+    "y_train = utils.get_labels(X_train_names)\n",
+    "y_test = utils.get_labels(X_test_names)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def load_features(names):\n",
+    "    output1 = [] # List of torso-arm vectors\n",
+    "    for filename in names:\n",
+    "        ps = load_ps('poses_compressed/frontraise/'+filename)\n",
+    "        poses = ps.poses\n",
+    "\n",
+    "        right_present = [1 for pose in poses \n",
+    "                if pose.rshoulder.exists and pose.relbow.exists and pose.rwrist.exists]\n",
+    "        left_present = [1 for pose in poses\n",
+    "                if pose.lshoulder.exists and pose.lelbow.exists and pose.lwrist.exists]\n",
+    "        right_count = sum(right_present)\n",
+    "        left_count = sum(left_present)\n",
+    "        side = 'right' if right_count > left_count else 'left'\n",
+    "\n",
+    "        if side == 'right':\n",
+    "            joints = [(pose.rshoulder, pose.relbow, pose.rwrist, pose.rhip, pose.neck) for pose in poses]\n",
+    "        else:\n",
+    "            joints = [(pose.lshoulder, pose.lelbow, pose.lwrist, pose.lhip, pose.neck) for pose in poses]\n",
+    "\n",
+    "        # filter out data points where a part does not exist\n",
+    "        joints = [joint for joint in joints if all(part.exists for part in joint)]\n",
+    "        joints = np.array(joints)\n",
+    "        \n",
+    "        # Shoulder to hip    \n",
+    "        torso_vecs = np.array([(joint[0].x - joint[3].x, joint[0].y - joint[3].y) for joint in joints])\n",
+    "        # Arm\n",
+    "        arm_vecs = np.array([(joint[0].x - joint[2].x, joint[0].y - joint[2].y) for joint in joints])\n",
+    "\n",
+    "        # normalize vectors\n",
+    "        torso_vecs = torso_vecs / np.expand_dims(np.linalg.norm(torso_vecs, axis=1), axis=1)\n",
+    "        arm_vecs = arm_vecs / np.expand_dims(np.linalg.norm(arm_vecs, axis=1), axis=1)\n",
+    "        \n",
+    "        angles = np.degrees(np.arccos(np.clip(np.sum(np.multiply(torso_vecs, arm_vecs), axis=1), -1.0, 1.0)))\n",
+    "        \n",
+    "        output1.append(angles.tolist())\n",
+    "    return output1\n",
+    "\n",
+    "X_train_1 = load_features(X_train_names)\n",
+    "X_test_1 = load_features(X_test_names)\n",
+    "#features = load_features(files)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "             precision    recall  f1-score   support\n",
+      "\n",
+      "    correct       1.00      1.00      1.00         6\n",
+      "  incorrect       1.00      1.00      1.00         6\n",
+      "\n",
+      "avg / total       1.00      1.00      1.00        12\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "predictions = []\n",
+    "for example in range(len(X_test_names)):\n",
+    "    # Store the average distance to good and bad training examples\n",
+    "    f1_good, f1_bad = [[] for i in range(2)]\n",
+    "    \n",
+    "    # Compare distance of current test example with all training examples\n",
+    "    for i in range(len(X_train_1)):\n",
+    "        dist1 = utils.DTWDistance(X_train_1[i], X_test_1[example])\n",
+    "        if y_train[i]:\n",
+    "            f1_good.append(dist1)\n",
+    "        else:\n",
+    "            f1_bad.append(dist1)\n",
+    "    good_score = np.mean(f1_good)\n",
+    "    bad_score = np.mean(f1_bad)\n",
+    "    if good_score < bad_score:\n",
+    "        predictions.append(1)\n",
+    "    else:\n",
+    "        predictions.append(0)\n",
+    "print(classification_report(y_test, predictions, target_names=['correct', 'incorrect']))"
+   ]
+  }
+ ],
+ "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": "3.5.4"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

ml_notebooks/DTW-and-KNN-ShoulderPress.ipynb → DTW-ShoulderPress.ipynb

@@ -27,7 +27,7 @@
    "outputs": [],
    "source": [
     "files = utils.files_in_order('poses_compressed/shoulderpress')\n",
-    "X_train_names, X_test_names = train_test_split(files, test_size=0.2, random_state=42)\n",
+    "X_train_names, X_test_names = train_test_split(files, test_size=0.4, random_state=42)\n",
     "y_train = utils.get_labels(X_train_names)\n",
     "y_test = utils.get_labels(X_test_names)"
    ]
@@ -103,10 +103,10 @@
      "text": [
       "             precision    recall  f1-score   support\n",
       "\n",
-      "    correct       0.20      0.33      0.25         3\n",
-      "  incorrect       0.33      0.20      0.25         5\n",
+      "    correct       0.67      0.86      0.75         7\n",
+      "  incorrect       0.83      0.62      0.71         8\n",
       "\n",
-      "avg / total       0.28      0.25      0.25         8\n",
+      "avg / total       0.76      0.73      0.73        15\n",
       "\n"
      ]
     }