{ "cells": [ { "cell_type": "code", "execution_count": 1, "id": "686b59bf", "metadata": {}, "outputs": [], "source": [ "import pandas as pd\n", "import numpy as np\n", "import matplotlib.pyplot as plt" ] }, { "cell_type": "code", "execution_count": 2, "id": "999530f1", "metadata": {}, "outputs": [], "source": [ "df = pd.read_csv(\"train_data.csv\")" ] }, { "cell_type": "code", "execution_count": 3, "id": "4f9a26dc", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
WRIST_xWRIST_yWRIST_zTHUMB_CMC_xTHUMB_CMC_yTHUMB_CMC_zTHUMB_MCP_xTHUMB_MCP_yTHUMB_MCP_zTHUMB_IP_x...PINKY_PIP_xPINKY_PIP_yPINKY_PIP_zPINKY_DIP_xPINKY_DIP_yPINKY_DIP_zPINKY_TIP_xPINKY_TIP_yPINKY_TIP_zlabel
00.00.00.0-0.057539-0.071465-0.026331-0.098735-0.197724-0.042757-0.068605...0.077761-0.316404-0.0828170.053253-0.244498-0.0735260.038338-0.190158-0.0587101.0
10.00.00.0-0.055692-0.070351-0.027606-0.095101-0.202065-0.043758-0.062707...0.087734-0.290525-0.0839750.059144-0.218611-0.0757230.040828-0.172123-0.0614421.0
20.00.00.0-0.055678-0.074819-0.025547-0.093986-0.207732-0.040555-0.060853...0.087272-0.296814-0.0823590.058133-0.222876-0.0742530.039232-0.172019-0.0600551.0
30.00.00.0-0.055489-0.070803-0.024467-0.095655-0.207985-0.038204-0.061606...0.087522-0.300159-0.0810850.058924-0.222508-0.0712550.042160-0.172443-0.0562751.0
40.00.00.0-0.056960-0.074619-0.021637-0.096975-0.209884-0.034189-0.064977...0.083821-0.303824-0.0796410.055938-0.228297-0.0700790.038268-0.174247-0.0553051.0
..................................................................
86890.00.00.00.047615-0.012006-0.0204680.084752-0.060672-0.0308240.084640...0.006213-0.271070-0.0487410.015103-0.318526-0.0550990.024375-0.362817-0.0600314.0
86900.00.00.00.047655-0.012923-0.0207660.085538-0.061602-0.0313130.084497...0.006827-0.270352-0.0482430.015869-0.318309-0.0544550.025217-0.363852-0.0594594.0
86910.00.00.00.048221-0.012413-0.0209680.087055-0.059296-0.0317420.085506...0.006929-0.267538-0.0486050.016013-0.315532-0.0550360.025359-0.361105-0.0602104.0
86920.00.00.00.047102-0.011459-0.0217080.085929-0.056868-0.0329440.084870...0.006963-0.267244-0.0481230.015571-0.314850-0.0544080.024368-0.359705-0.0594074.0
86930.00.00.00.046908-0.011728-0.0211160.085699-0.058851-0.0320610.085546...0.006706-0.267916-0.0498470.015464-0.316603-0.0562890.024499-0.362343-0.0614624.0
\n", "

8694 rows × 64 columns

\n", "
" ], "text/plain": [ " WRIST_x WRIST_y WRIST_z THUMB_CMC_x THUMB_CMC_y THUMB_CMC_z \\\n", "0 0.0 0.0 0.0 -0.057539 -0.071465 -0.026331 \n", "1 0.0 0.0 0.0 -0.055692 -0.070351 -0.027606 \n", "2 0.0 0.0 0.0 -0.055678 -0.074819 -0.025547 \n", "3 0.0 0.0 0.0 -0.055489 -0.070803 -0.024467 \n", "4 0.0 0.0 0.0 -0.056960 -0.074619 -0.021637 \n", "... ... ... ... ... ... ... \n", "8689 0.0 0.0 0.0 0.047615 -0.012006 -0.020468 \n", "8690 0.0 0.0 0.0 0.047655 -0.012923 -0.020766 \n", "8691 0.0 0.0 0.0 0.048221 -0.012413 -0.020968 \n", "8692 0.0 0.0 0.0 0.047102 -0.011459 -0.021708 \n", "8693 0.0 0.0 0.0 0.046908 -0.011728 -0.021116 \n", "\n", " THUMB_MCP_x THUMB_MCP_y THUMB_MCP_z THUMB_IP_x ... PINKY_PIP_x \\\n", "0 -0.098735 -0.197724 -0.042757 -0.068605 ... 0.077761 \n", "1 -0.095101 -0.202065 -0.043758 -0.062707 ... 0.087734 \n", "2 -0.093986 -0.207732 -0.040555 -0.060853 ... 0.087272 \n", "3 -0.095655 -0.207985 -0.038204 -0.061606 ... 0.087522 \n", "4 -0.096975 -0.209884 -0.034189 -0.064977 ... 0.083821 \n", "... ... ... ... ... ... ... \n", "8689 0.084752 -0.060672 -0.030824 0.084640 ... 0.006213 \n", "8690 0.085538 -0.061602 -0.031313 0.084497 ... 0.006827 \n", "8691 0.087055 -0.059296 -0.031742 0.085506 ... 0.006929 \n", "8692 0.085929 -0.056868 -0.032944 0.084870 ... 0.006963 \n", "8693 0.085699 -0.058851 -0.032061 0.085546 ... 0.006706 \n", "\n", " PINKY_PIP_y PINKY_PIP_z PINKY_DIP_x PINKY_DIP_y PINKY_DIP_z \\\n", "0 -0.316404 -0.082817 0.053253 -0.244498 -0.073526 \n", "1 -0.290525 -0.083975 0.059144 -0.218611 -0.075723 \n", "2 -0.296814 -0.082359 0.058133 -0.222876 -0.074253 \n", "3 -0.300159 -0.081085 0.058924 -0.222508 -0.071255 \n", "4 -0.303824 -0.079641 0.055938 -0.228297 -0.070079 \n", "... ... ... ... ... ... \n", "8689 -0.271070 -0.048741 0.015103 -0.318526 -0.055099 \n", "8690 -0.270352 -0.048243 0.015869 -0.318309 -0.054455 \n", "8691 -0.267538 -0.048605 0.016013 -0.315532 -0.055036 \n", "8692 -0.267244 -0.048123 0.015571 -0.314850 -0.054408 \n", "8693 -0.267916 -0.049847 0.015464 -0.316603 -0.056289 \n", "\n", " PINKY_TIP_x PINKY_TIP_y PINKY_TIP_z label \n", "0 0.038338 -0.190158 -0.058710 1.0 \n", "1 0.040828 -0.172123 -0.061442 1.0 \n", "2 0.039232 -0.172019 -0.060055 1.0 \n", "3 0.042160 -0.172443 -0.056275 1.0 \n", "4 0.038268 -0.174247 -0.055305 1.0 \n", "... ... ... ... ... \n", "8689 0.024375 -0.362817 -0.060031 4.0 \n", "8690 0.025217 -0.363852 -0.059459 4.0 \n", "8691 0.025359 -0.361105 -0.060210 4.0 \n", "8692 0.024368 -0.359705 -0.059407 4.0 \n", "8693 0.024499 -0.362343 -0.061462 4.0 \n", "\n", "[8694 rows x 64 columns]" ] }, "execution_count": 3, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df" ] }, { "cell_type": "code", "execution_count": 4, "id": "8c13c069", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
WRIST_xWRIST_yWRIST_zTHUMB_CMC_xTHUMB_CMC_yTHUMB_CMC_zTHUMB_MCP_xTHUMB_MCP_yTHUMB_MCP_zTHUMB_IP_x...PINKY_PIP_xPINKY_PIP_yPINKY_PIP_zPINKY_DIP_xPINKY_DIP_yPINKY_DIP_zPINKY_TIP_xPINKY_TIP_yPINKY_TIP_zlabel
count8694.08694.08694.08694.0000008694.0000008694.0000008694.0000008694.0000008694.0000008694.000000...8694.0000008694.0000008694.0000008694.0000008694.0000008694.0000008694.0000008694.0000008694.0000008694.000000
mean0.00.00.00.012662-0.049743-0.0158740.022310-0.135511-0.0250030.021180...-0.004829-0.296526-0.059288-0.006298-0.305757-0.062451-0.007505-0.316906-0.0625663.114217
std0.00.00.00.0509080.0354130.0137020.0894820.0526590.0183010.094838...0.0634360.0663940.0206120.0677090.1134670.0237060.0714960.1608320.0275971.207629
min0.00.00.0-0.086818-0.173453-0.065661-0.156700-0.319774-0.110080-0.209468...-0.228817-0.543269-0.140484-0.253123-0.641913-0.155522-0.273237-0.745557-0.1701661.000000
25%0.00.00.0-0.049603-0.058369-0.023474-0.082634-0.164581-0.035496-0.066085...-0.039343-0.339477-0.072416-0.041956-0.397840-0.076801-0.044047-0.453859-0.0789512.000000
50%0.00.00.00.042084-0.038699-0.0186050.066347-0.125283-0.0275820.042749...-0.004672-0.294442-0.059133-0.005029-0.313036-0.062971-0.005147-0.353151-0.0628063.000000
75%0.00.00.00.053208-0.025961-0.0123820.096309-0.097576-0.0183790.097657...0.028778-0.245439-0.0468090.028609-0.196222-0.0487100.028313-0.152212-0.0460694.000000
max0.00.00.00.1000040.0158630.0648520.183811-0.0097140.0991120.209236...0.219966-0.0996150.0664330.246374-0.0671620.0756930.282817-0.0640960.0827605.000000
\n", "

8 rows × 64 columns

\n", "
" ], "text/plain": [ " WRIST_x WRIST_y WRIST_z THUMB_CMC_x THUMB_CMC_y THUMB_CMC_z \\\n", "count 8694.0 8694.0 8694.0 8694.000000 8694.000000 8694.000000 \n", "mean 0.0 0.0 0.0 0.012662 -0.049743 -0.015874 \n", "std 0.0 0.0 0.0 0.050908 0.035413 0.013702 \n", "min 0.0 0.0 0.0 -0.086818 -0.173453 -0.065661 \n", "25% 0.0 0.0 0.0 -0.049603 -0.058369 -0.023474 \n", "50% 0.0 0.0 0.0 0.042084 -0.038699 -0.018605 \n", "75% 0.0 0.0 0.0 0.053208 -0.025961 -0.012382 \n", "max 0.0 0.0 0.0 0.100004 0.015863 0.064852 \n", "\n", " THUMB_MCP_x THUMB_MCP_y THUMB_MCP_z THUMB_IP_x ... PINKY_PIP_x \\\n", "count 8694.000000 8694.000000 8694.000000 8694.000000 ... 8694.000000 \n", "mean 0.022310 -0.135511 -0.025003 0.021180 ... -0.004829 \n", "std 0.089482 0.052659 0.018301 0.094838 ... 0.063436 \n", "min -0.156700 -0.319774 -0.110080 -0.209468 ... -0.228817 \n", "25% -0.082634 -0.164581 -0.035496 -0.066085 ... -0.039343 \n", "50% 0.066347 -0.125283 -0.027582 0.042749 ... -0.004672 \n", "75% 0.096309 -0.097576 -0.018379 0.097657 ... 0.028778 \n", "max 0.183811 -0.009714 0.099112 0.209236 ... 0.219966 \n", "\n", " PINKY_PIP_y PINKY_PIP_z PINKY_DIP_x PINKY_DIP_y PINKY_DIP_z \\\n", "count 8694.000000 8694.000000 8694.000000 8694.000000 8694.000000 \n", "mean -0.296526 -0.059288 -0.006298 -0.305757 -0.062451 \n", "std 0.066394 0.020612 0.067709 0.113467 0.023706 \n", "min -0.543269 -0.140484 -0.253123 -0.641913 -0.155522 \n", "25% -0.339477 -0.072416 -0.041956 -0.397840 -0.076801 \n", "50% -0.294442 -0.059133 -0.005029 -0.313036 -0.062971 \n", "75% -0.245439 -0.046809 0.028609 -0.196222 -0.048710 \n", "max -0.099615 0.066433 0.246374 -0.067162 0.075693 \n", "\n", " PINKY_TIP_x PINKY_TIP_y PINKY_TIP_z label \n", "count 8694.000000 8694.000000 8694.000000 8694.000000 \n", "mean -0.007505 -0.316906 -0.062566 3.114217 \n", "std 0.071496 0.160832 0.027597 1.207629 \n", "min -0.273237 -0.745557 -0.170166 1.000000 \n", "25% -0.044047 -0.453859 -0.078951 2.000000 \n", "50% -0.005147 -0.353151 -0.062806 3.000000 \n", "75% 0.028313 -0.152212 -0.046069 4.000000 \n", "max 0.282817 -0.064096 0.082760 5.000000 \n", "\n", "[8 rows x 64 columns]" ] }, "execution_count": 4, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df.describe()" ] }, { "cell_type": "code", "execution_count": 5, "id": "d042c994", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "\n", "RangeIndex: 8694 entries, 0 to 8693\n", "Data columns (total 64 columns):\n", " # Column Non-Null Count Dtype \n", "--- ------ -------------- ----- \n", " 0 WRIST_x 8694 non-null float64\n", " 1 WRIST_y 8694 non-null float64\n", " 2 WRIST_z 8694 non-null float64\n", " 3 THUMB_CMC_x 8694 non-null float64\n", " 4 THUMB_CMC_y 8694 non-null float64\n", " 5 THUMB_CMC_z 8694 non-null float64\n", " 6 THUMB_MCP_x 8694 non-null float64\n", " 7 THUMB_MCP_y 8694 non-null float64\n", " 8 THUMB_MCP_z 8694 non-null float64\n", " 9 THUMB_IP_x 8694 non-null float64\n", " 10 THUMB_IP_y 8694 non-null float64\n", " 11 THUMB_IP_z 8694 non-null float64\n", " 12 THUMB_TIP_x 8694 non-null float64\n", " 13 THUMB_TIP_y 8694 non-null float64\n", " 14 THUMB_TIP_z 8694 non-null float64\n", " 15 INDEX_FINGER_MCP_x 8694 non-null float64\n", " 16 INDEX_FINGER_MCP_y 8694 non-null float64\n", " 17 INDEX_FINGER_MCP_z 8694 non-null float64\n", " 18 INDEX_FINGER_PIP_x 8694 non-null float64\n", " 19 INDEX_FINGER_PIP_y 8694 non-null float64\n", " 20 INDEX_FINGER_PIP_z 8694 non-null float64\n", " 21 INDEX_FINGER_DIP_x 8694 non-null float64\n", " 22 INDEX_FINGER_DIP_y 8694 non-null float64\n", " 23 INDEX_FINGER_DIP_z 8694 non-null float64\n", " 24 INDEX_FINGER_TIP_x 8694 non-null float64\n", " 25 INDEX_FINGER_TIP_y 8694 non-null float64\n", " 26 INDEX_FINGER_TIP_z 8694 non-null float64\n", " 27 MIDDLE_FINGER_MCP_x 8694 non-null float64\n", " 28 MIDDLE_FINGER_MCP_y 8694 non-null float64\n", " 29 MIDDLE_FINGER_MCP_z 8694 non-null float64\n", " 30 MIDDLE_FINGER_PIP_x 8694 non-null float64\n", " 31 MIDDLE_FINGER_PIP_y 8694 non-null float64\n", " 32 MIDDLE_FINGER_PIP_z 8694 non-null float64\n", " 33 MIDDLE_FINGER_DIP_x 8694 non-null float64\n", " 34 MIDDLE_FINGER_DIP_y 8694 non-null float64\n", " 35 MIDDLE_FINGER_DIP_z 8694 non-null float64\n", " 36 MIDDLE_FINGER_TIP_x 8694 non-null float64\n", " 37 MIDDLE_FINGER_TIP_y 8694 non-null float64\n", " 38 MIDDLE_FINGER_TIP_z 8694 non-null float64\n", " 39 RING_FINGER_MCP_x 8694 non-null float64\n", " 40 RING_FINGER_MCP_y 8694 non-null float64\n", " 41 RING_FINGER_MCP_z 8694 non-null float64\n", " 42 RING_FINGER_PIP_x 8694 non-null float64\n", " 43 RING_FINGER_PIP_y 8694 non-null float64\n", " 44 RING_FINGER_PIP_z 8694 non-null float64\n", " 45 RING_FINGER_DIP_x 8694 non-null float64\n", " 46 RING_FINGER_DIP_y 8694 non-null float64\n", " 47 RING_FINGER_DIP_z 8694 non-null float64\n", " 48 RING_FINGER_TIP_x 8694 non-null float64\n", " 49 RING_FINGER_TIP_y 8694 non-null float64\n", " 50 RING_FINGER_TIP_z 8694 non-null float64\n", " 51 PINKY_MCP_x 8694 non-null float64\n", " 52 PINKY_MCP_y 8694 non-null float64\n", " 53 PINKY_MCP_z 8694 non-null float64\n", " 54 PINKY_PIP_x 8694 non-null float64\n", " 55 PINKY_PIP_y 8694 non-null float64\n", " 56 PINKY_PIP_z 8694 non-null float64\n", " 57 PINKY_DIP_x 8694 non-null float64\n", " 58 PINKY_DIP_y 8694 non-null float64\n", " 59 PINKY_DIP_z 8694 non-null float64\n", " 60 PINKY_TIP_x 8694 non-null float64\n", " 61 PINKY_TIP_y 8694 non-null float64\n", " 62 PINKY_TIP_z 8694 non-null float64\n", " 63 label 8694 non-null float64\n", "dtypes: float64(64)\n", "memory usage: 4.2 MB\n" ] } ], "source": [ "df.info()" ] }, { "cell_type": "code", "execution_count": 6, "id": "0bcacdc1", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "label\n", "1.0 1064\n", "2.0 1424\n", "3.0 2944\n", "4.0 1979\n", "5.0 1283\n", "Name: WRIST_x, dtype: int64" ] }, "execution_count": 6, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df.groupby(\"label\")[\"WRIST_x\"].count()" ] }, { "cell_type": "markdown", "id": "52694bfb", "metadata": {}, "source": [ "..One time fix..\n", "I need to fix the old data and subtract the wrist coordinates from each features " ] }, { "cell_type": "code", "execution_count": 7, "id": "886f7270", "metadata": {}, "outputs": [], "source": [ "def func(row):\n", " x, y, z = row.iloc[0], row.iloc[1], row.iloc[2] # wrist xyz\n", " for i in range(0, len(row)-1, 3):\n", " row.iloc[i] = row.iloc[i] - x\n", " for i in range(1, len(row)-1, 3):\n", " row.iloc[i] = row.iloc[i] - y\n", " for i in range(2, len(row)-1, 3):\n", " row.iloc[i] = row.iloc[i] - z\n", " return row\n", "\n", "# df = df.apply(func, axis=1)\n", "# df" ] }, { "cell_type": "code", "execution_count": 8, "id": "a2ef013f", "metadata": {}, "outputs": [], "source": [ "# df.to_csv('train_data.csv', index=False)" ] }, { "cell_type": "code", "execution_count": 9, "id": "8a7c059d", "metadata": {}, "outputs": [], "source": [ "from sklearn.model_selection import train_test_split\n", "\n", "train_set, test_set = train_test_split(df, test_size=0.3, random_state=42, shuffle=True, stratify=df['label'])" ] }, { "cell_type": "code", "execution_count": 10, "id": "1801ed2c", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
WRIST_xWRIST_yWRIST_zTHUMB_CMC_xTHUMB_CMC_yTHUMB_CMC_zTHUMB_MCP_xTHUMB_MCP_yTHUMB_MCP_zTHUMB_IP_x...PINKY_PIP_xPINKY_PIP_yPINKY_PIP_zPINKY_DIP_xPINKY_DIP_yPINKY_DIP_zPINKY_TIP_xPINKY_TIP_yPINKY_TIP_zlabel
20960.00.00.0-0.031134-0.1279670.003506-0.021370-0.229542-0.0073580.016917...0.110509-0.236130-0.0597500.091704-0.211687-0.0430230.080731-0.181086-0.0280282.0
57410.00.00.0-0.025325-0.1194080.018037-0.017356-0.2148790.0221070.010074...0.118601-0.274658-0.0527180.139149-0.323255-0.0584640.156074-0.366578-0.0610984.0
27640.00.00.00.034237-0.0852080.0116720.044016-0.1560970.0121280.029640...-0.103818-0.283978-0.037229-0.117065-0.331457-0.039452-0.127330-0.374249-0.0409783.0
25240.00.00.0-0.064078-0.049834-0.013207-0.114952-0.152806-0.025534-0.144391...0.070130-0.403020-0.0597930.079705-0.473826-0.0609020.083696-0.544319-0.0612023.0
84490.00.00.00.046406-0.019666-0.0177730.082191-0.079571-0.0238870.078505...0.001426-0.273189-0.0287440.008954-0.319018-0.0331540.016512-0.361284-0.0364514.0
..................................................................
8090.00.00.00.052863-0.032350-0.0249390.099440-0.110987-0.0395550.086531...-0.028137-0.238647-0.070566-0.018157-0.172572-0.060498-0.014315-0.134769-0.0457411.0
64340.00.00.00.029594-0.106317-0.0219150.045134-0.209189-0.0385520.062114...-0.163283-0.135134-0.052690-0.192641-0.149463-0.062431-0.217876-0.167613-0.0678765.0
76280.00.00.00.039779-0.034610-0.0122000.063473-0.103854-0.0132790.042186...0.002722-0.238822-0.0363320.010146-0.194895-0.0382730.013473-0.144082-0.0373703.0
85190.00.00.00.055265-0.020795-0.0219060.097875-0.090066-0.0298290.093866...-0.002164-0.333022-0.0397090.007218-0.387862-0.0451320.017621-0.437138-0.0490994.0
53580.00.00.00.049662-0.046963-0.0297520.084285-0.123904-0.0475840.050442...-0.084367-0.297632-0.096556-0.110932-0.345592-0.111425-0.133137-0.396410-0.1207494.0
\n", "

6085 rows × 64 columns

\n", "
" ], "text/plain": [ " WRIST_x WRIST_y WRIST_z THUMB_CMC_x THUMB_CMC_y THUMB_CMC_z \\\n", "2096 0.0 0.0 0.0 -0.031134 -0.127967 0.003506 \n", "5741 0.0 0.0 0.0 -0.025325 -0.119408 0.018037 \n", "2764 0.0 0.0 0.0 0.034237 -0.085208 0.011672 \n", "2524 0.0 0.0 0.0 -0.064078 -0.049834 -0.013207 \n", "8449 0.0 0.0 0.0 0.046406 -0.019666 -0.017773 \n", "... ... ... ... ... ... ... \n", "809 0.0 0.0 0.0 0.052863 -0.032350 -0.024939 \n", "6434 0.0 0.0 0.0 0.029594 -0.106317 -0.021915 \n", "7628 0.0 0.0 0.0 0.039779 -0.034610 -0.012200 \n", "8519 0.0 0.0 0.0 0.055265 -0.020795 -0.021906 \n", "5358 0.0 0.0 0.0 0.049662 -0.046963 -0.029752 \n", "\n", " THUMB_MCP_x THUMB_MCP_y THUMB_MCP_z THUMB_IP_x ... PINKY_PIP_x \\\n", "2096 -0.021370 -0.229542 -0.007358 0.016917 ... 0.110509 \n", "5741 -0.017356 -0.214879 0.022107 0.010074 ... 0.118601 \n", "2764 0.044016 -0.156097 0.012128 0.029640 ... -0.103818 \n", "2524 -0.114952 -0.152806 -0.025534 -0.144391 ... 0.070130 \n", "8449 0.082191 -0.079571 -0.023887 0.078505 ... 0.001426 \n", "... ... ... ... ... ... ... \n", "809 0.099440 -0.110987 -0.039555 0.086531 ... -0.028137 \n", "6434 0.045134 -0.209189 -0.038552 0.062114 ... -0.163283 \n", "7628 0.063473 -0.103854 -0.013279 0.042186 ... 0.002722 \n", "8519 0.097875 -0.090066 -0.029829 0.093866 ... -0.002164 \n", "5358 0.084285 -0.123904 -0.047584 0.050442 ... -0.084367 \n", "\n", " PINKY_PIP_y PINKY_PIP_z PINKY_DIP_x PINKY_DIP_y PINKY_DIP_z \\\n", "2096 -0.236130 -0.059750 0.091704 -0.211687 -0.043023 \n", "5741 -0.274658 -0.052718 0.139149 -0.323255 -0.058464 \n", "2764 -0.283978 -0.037229 -0.117065 -0.331457 -0.039452 \n", "2524 -0.403020 -0.059793 0.079705 -0.473826 -0.060902 \n", "8449 -0.273189 -0.028744 0.008954 -0.319018 -0.033154 \n", "... ... ... ... ... ... \n", "809 -0.238647 -0.070566 -0.018157 -0.172572 -0.060498 \n", "6434 -0.135134 -0.052690 -0.192641 -0.149463 -0.062431 \n", "7628 -0.238822 -0.036332 0.010146 -0.194895 -0.038273 \n", "8519 -0.333022 -0.039709 0.007218 -0.387862 -0.045132 \n", "5358 -0.297632 -0.096556 -0.110932 -0.345592 -0.111425 \n", "\n", " PINKY_TIP_x PINKY_TIP_y PINKY_TIP_z label \n", "2096 0.080731 -0.181086 -0.028028 2.0 \n", "5741 0.156074 -0.366578 -0.061098 4.0 \n", "2764 -0.127330 -0.374249 -0.040978 3.0 \n", "2524 0.083696 -0.544319 -0.061202 3.0 \n", "8449 0.016512 -0.361284 -0.036451 4.0 \n", "... ... ... ... ... \n", "809 -0.014315 -0.134769 -0.045741 1.0 \n", "6434 -0.217876 -0.167613 -0.067876 5.0 \n", "7628 0.013473 -0.144082 -0.037370 3.0 \n", "8519 0.017621 -0.437138 -0.049099 4.0 \n", "5358 -0.133137 -0.396410 -0.120749 4.0 \n", "\n", "[6085 rows x 64 columns]" ] }, "execution_count": 10, "metadata": {}, "output_type": "execute_result" } ], "source": [ "train_set" ] }, { "cell_type": "code", "execution_count": 11, "id": "06a3dba6", "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
WRIST_xWRIST_yWRIST_zTHUMB_CMC_xTHUMB_CMC_yTHUMB_CMC_zTHUMB_MCP_xTHUMB_MCP_yTHUMB_MCP_zTHUMB_IP_x...PINKY_PIP_xPINKY_PIP_yPINKY_PIP_zPINKY_DIP_xPINKY_DIP_yPINKY_DIP_zPINKY_TIP_xPINKY_TIP_yPINKY_TIP_zlabel
8260.00.00.00.048946-0.025729-0.0223600.091728-0.101759-0.0348670.086899...-0.031575-0.228799-0.066676-0.019560-0.166227-0.058096-0.013957-0.121445-0.0440541.0
37890.00.00.00.051196-0.039587-0.0202310.087972-0.115390-0.0302980.101004...-0.032819-0.319158-0.048365-0.029740-0.376926-0.051765-0.025546-0.432299-0.0537313.0
2190.00.00.00.053897-0.0809580.0009280.078084-0.172383-0.0045190.056157...-0.034718-0.236302-0.027854-0.026203-0.206872-0.015052-0.027840-0.180848-0.0061551.0
8310.00.00.00.064432-0.036300-0.0304730.117950-0.140157-0.0469470.094093...-0.041053-0.276031-0.087501-0.025712-0.196638-0.076604-0.019014-0.153718-0.0599961.0
30380.00.00.0-0.057135-0.023739-0.011102-0.108559-0.119451-0.017151-0.110449...0.044515-0.367976-0.0539280.054518-0.430534-0.0598180.059544-0.487775-0.0641853.0
..................................................................
2300.00.00.0-0.046368-0.082749-0.009246-0.082470-0.189178-0.020109-0.079587...0.030542-0.273261-0.0634650.010343-0.210576-0.0551500.010728-0.159244-0.0432141.0
55820.00.00.00.045595-0.027506-0.0208060.084642-0.090396-0.0310990.073277...-0.023350-0.310820-0.059181-0.028658-0.370852-0.069567-0.032117-0.426881-0.0762894.0
42020.00.00.00.044418-0.024597-0.0160890.077327-0.107340-0.0209560.057450...-0.011450-0.254861-0.0696540.001848-0.194567-0.0836510.010194-0.134239-0.0920973.0
78580.00.00.00.044177-0.020785-0.0429680.070827-0.099605-0.0592450.033620...-0.022348-0.275270-0.088156-0.013294-0.208049-0.101394-0.008838-0.149454-0.1064273.0
54430.00.00.00.053325-0.036923-0.0266810.081927-0.118427-0.0411820.055795...-0.013200-0.380243-0.056703-0.014108-0.444517-0.063443-0.013874-0.501148-0.0678564.0
\n", "

2609 rows × 64 columns

\n", "
" ], "text/plain": [ " WRIST_x WRIST_y WRIST_z THUMB_CMC_x THUMB_CMC_y THUMB_CMC_z \\\n", "826 0.0 0.0 0.0 0.048946 -0.025729 -0.022360 \n", "3789 0.0 0.0 0.0 0.051196 -0.039587 -0.020231 \n", "219 0.0 0.0 0.0 0.053897 -0.080958 0.000928 \n", "831 0.0 0.0 0.0 0.064432 -0.036300 -0.030473 \n", "3038 0.0 0.0 0.0 -0.057135 -0.023739 -0.011102 \n", "... ... ... ... ... ... ... \n", "230 0.0 0.0 0.0 -0.046368 -0.082749 -0.009246 \n", "5582 0.0 0.0 0.0 0.045595 -0.027506 -0.020806 \n", "4202 0.0 0.0 0.0 0.044418 -0.024597 -0.016089 \n", "7858 0.0 0.0 0.0 0.044177 -0.020785 -0.042968 \n", "5443 0.0 0.0 0.0 0.053325 -0.036923 -0.026681 \n", "\n", " THUMB_MCP_x THUMB_MCP_y THUMB_MCP_z THUMB_IP_x ... PINKY_PIP_x \\\n", "826 0.091728 -0.101759 -0.034867 0.086899 ... -0.031575 \n", "3789 0.087972 -0.115390 -0.030298 0.101004 ... -0.032819 \n", "219 0.078084 -0.172383 -0.004519 0.056157 ... -0.034718 \n", "831 0.117950 -0.140157 -0.046947 0.094093 ... -0.041053 \n", "3038 -0.108559 -0.119451 -0.017151 -0.110449 ... 0.044515 \n", "... ... ... ... ... ... ... \n", "230 -0.082470 -0.189178 -0.020109 -0.079587 ... 0.030542 \n", "5582 0.084642 -0.090396 -0.031099 0.073277 ... -0.023350 \n", "4202 0.077327 -0.107340 -0.020956 0.057450 ... -0.011450 \n", "7858 0.070827 -0.099605 -0.059245 0.033620 ... -0.022348 \n", "5443 0.081927 -0.118427 -0.041182 0.055795 ... -0.013200 \n", "\n", " PINKY_PIP_y PINKY_PIP_z PINKY_DIP_x PINKY_DIP_y PINKY_DIP_z \\\n", "826 -0.228799 -0.066676 -0.019560 -0.166227 -0.058096 \n", "3789 -0.319158 -0.048365 -0.029740 -0.376926 -0.051765 \n", "219 -0.236302 -0.027854 -0.026203 -0.206872 -0.015052 \n", "831 -0.276031 -0.087501 -0.025712 -0.196638 -0.076604 \n", "3038 -0.367976 -0.053928 0.054518 -0.430534 -0.059818 \n", "... ... ... ... ... ... \n", "230 -0.273261 -0.063465 0.010343 -0.210576 -0.055150 \n", "5582 -0.310820 -0.059181 -0.028658 -0.370852 -0.069567 \n", "4202 -0.254861 -0.069654 0.001848 -0.194567 -0.083651 \n", "7858 -0.275270 -0.088156 -0.013294 -0.208049 -0.101394 \n", "5443 -0.380243 -0.056703 -0.014108 -0.444517 -0.063443 \n", "\n", " PINKY_TIP_x PINKY_TIP_y PINKY_TIP_z label \n", "826 -0.013957 -0.121445 -0.044054 1.0 \n", "3789 -0.025546 -0.432299 -0.053731 3.0 \n", "219 -0.027840 -0.180848 -0.006155 1.0 \n", "831 -0.019014 -0.153718 -0.059996 1.0 \n", "3038 0.059544 -0.487775 -0.064185 3.0 \n", "... ... ... ... ... \n", "230 0.010728 -0.159244 -0.043214 1.0 \n", "5582 -0.032117 -0.426881 -0.076289 4.0 \n", "4202 0.010194 -0.134239 -0.092097 3.0 \n", "7858 -0.008838 -0.149454 -0.106427 3.0 \n", "5443 -0.013874 -0.501148 -0.067856 4.0 \n", "\n", "[2609 rows x 64 columns]" ] }, "execution_count": 11, "metadata": {}, "output_type": "execute_result" } ], "source": [ "test_set" ] }, { "cell_type": "code", "execution_count": 12, "id": "2ece1264", "metadata": {}, "outputs": [], "source": [ "y_train = train_set[['label']] - 1 #torch dataset expects 0 indexing\n", "x_train = train_set.copy()\n", "del x_train['label']\n", "\n", "y_test = test_set[['label']] - 1\n", "x_test = test_set.copy()\n", "del x_test['label']" ] }, { "cell_type": "code", "execution_count": 13, "id": "00ef789b", "metadata": {}, "outputs": [], "source": [ "import torch\n", "from torch.utils.data import DataLoader, TensorDataset\n", "import torch.nn as nn" ] }, { "cell_type": "code", "execution_count": 14, "id": "69a55d33", "metadata": {}, "outputs": [], "source": [ "x_train = torch.tensor(x_train.values.astype(np.float32))\n", "y_train = torch.tensor(y_train.values.astype(np.int32)).squeeze()\n", "x_test = torch.tensor(x_test.values.astype(np.float32))\n", "y_test = torch.tensor(y_test.values.astype(np.int32)).squeeze()\n", "\n", "train_dataset = TensorDataset(x_train, y_train)\n", "test_dataset = TensorDataset(x_test, y_test)" ] }, { "cell_type": "code", "execution_count": 15, "id": "cf3eac06", "metadata": {}, "outputs": [], "source": [ "class TransformDataset(torch.utils.data.Dataset):\n", " def __init__(self, dataset, transform=None):\n", " self.dataset = dataset\n", " self.transform=transform\n", " def __len__(self):\n", " return len(self.dataset)\n", " def __getitem__(self, index):\n", " x, y = self.dataset[index]\n", " if self.transform:\n", " x = self.transform(x)\n", " return x, y" ] }, { "cell_type": "code", "execution_count": 16, "id": "c8120154", "metadata": {}, "outputs": [], "source": [ "import torch\n", "def transformations(x):\n", " noise = torch.randn_like(x) * 0.02\n", " scale = torch.empty(1, device=x.device).uniform_(0.98, 1.02)\n", " return (x + noise) * scale" ] }, { "cell_type": "code", "execution_count": 17, "id": "81ed42fc", "metadata": {}, "outputs": [], "source": [ "train_dataset = TransformDataset(train_dataset, transform=transformations)" ] }, { "cell_type": "code", "execution_count": 18, "id": "095f9061", "metadata": {}, "outputs": [], "source": [ "train_loader = DataLoader(train_dataset, 64, shuffle=True, num_workers=0)\n", "test_loader = DataLoader(test_dataset, 64, shuffle=False, num_workers=0)" ] }, { "cell_type": "code", "execution_count": 19, "id": "93dd6251", "metadata": {}, "outputs": [], "source": [ "class Classifier(nn.Module):\n", " def __init__(self):\n", " super().__init__()\n", " \n", " self.feedforward = nn.Sequential(\n", " nn.BatchNorm1d(63),\n", " nn.Linear(in_features=63, out_features=256),\n", " nn.ReLU(),\n", " nn.Dropout(0.3),\n", " \n", " nn.BatchNorm1d(256),\n", " nn.Linear(in_features=256, out_features=128),\n", " nn.ReLU(),\n", " nn.Dropout(0.3),\n", " \n", " nn.BatchNorm1d(128),\n", " nn.Linear(in_features=128, out_features=5),\n", " )\n", " \n", " def forward(self, x):\n", " x = self.feedforward(x)\n", " return x" ] }, { "cell_type": "code", "execution_count": 20, "id": "12ce0dcd", "metadata": {}, "outputs": [], "source": [ "device = 'mps' if torch.mps.is_available() else 'cpu'" ] }, { "cell_type": "code", "execution_count": 21, "id": "73ae6ccb", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "OptimizedModule(\n", " (_orig_mod): Classifier(\n", " (feedforward): Sequential(\n", " (0): BatchNorm1d(63, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", " (1): Linear(in_features=63, out_features=256, bias=True)\n", " (2): ReLU()\n", " (3): Dropout(p=0.3, inplace=False)\n", " (4): BatchNorm1d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", " (5): Linear(in_features=256, out_features=128, bias=True)\n", " (6): ReLU()\n", " (7): Dropout(p=0.3, inplace=False)\n", " (8): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", " (9): Linear(in_features=128, out_features=5, bias=True)\n", " )\n", " )\n", ")" ] }, "execution_count": 21, "metadata": {}, "output_type": "execute_result" } ], "source": [ "model = Classifier().to(device)\n", "torch.compile(model)" ] }, { "cell_type": "code", "execution_count": 22, "id": "73a78f2d", "metadata": {}, "outputs": [], "source": [ "loss_fn = nn.CrossEntropyLoss()\n", "optimizer = torch.optim.AdamW(model.parameters(), lr=0.005, weight_decay=0.01)\n", "scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.5)" ] }, { "cell_type": "code", "execution_count": 23, "id": "e2f43127", "metadata": {}, "outputs": [], "source": [ "def get_accuracy():\n", " model.eval()\n", " with torch.no_grad():\n", " total_correct = 0\n", " for X, y in test_loader:\n", " X, y = X.to(device), y.to(device)\n", " \n", " output = model(X)\n", " total_correct += (torch.argmax(output, dim=1) == y).sum().item()\n", " \n", " test_acc = total_correct / len(test_loader.dataset)\n", " return test_acc\n" ] }, { "cell_type": "code", "execution_count": 24, "id": "5e04ea98", "metadata": {}, "outputs": [], "source": [ "def train(epochs:int):\n", " \n", " for epoch in range(epochs):\n", " model.train()\n", " epoch_loss = 0\n", " \n", " total_correct = 0\n", " \n", " for X, y in train_loader:\n", " X, y = X.to(device), y.to(device)\n", " \n", " optimizer.zero_grad()\n", " \n", " output = model(X)\n", " \n", " total_correct += (torch.argmax(output, dim=1) == y).sum().item()\n", " \n", " loss = loss_fn(output, y)\n", " epoch_loss += loss.item()\n", " loss.backward()\n", " \n", " optimizer.step()\n", " \n", " scheduler.step()\n", " train_acc = total_correct / len(train_loader.dataset)\n", " \n", " print(f\"Epoch {epoch+1}: loss: {epoch_loss / len(train_loader)}, Train Accuracy: {train_acc}, Test Accuracy: {get_accuracy()}, LR = {optimizer.param_groups[0]['lr']}\")" ] }, { "cell_type": "code", "execution_count": 25, "id": "640f3cc0", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Epoch 1: loss: 0.27147899529275793, Train Accuracy: 0.9089564502875924, Test Accuracy: 0.9900344959754696, LR = 0.005\n", "Epoch 2: loss: 0.1263906546907189, Train Accuracy: 0.9654889071487264, Test Accuracy: 0.9931008049060943, LR = 0.005\n", "Epoch 3: loss: 0.08338426114642061, Train Accuracy: 0.9788003286770748, Test Accuracy: 0.9900344959754696, LR = 0.005\n", "Epoch 4: loss: 0.07302799479415019, Train Accuracy: 0.9797863599013968, Test Accuracy: 0.9931008049060943, LR = 0.005\n", "Epoch 5: loss: 0.09298191487808556, Train Accuracy: 0.9756778964667214, Test Accuracy: 0.9911843618244538, LR = 0.005\n", "Epoch 6: loss: 0.08420706538405891, Train Accuracy: 0.9750205423171734, Test Accuracy: 0.9934840935224224, LR = 0.005\n", "Epoch 7: loss: 0.06479617586107149, Train Accuracy: 0.9824157764995891, Test Accuracy: 0.9938673821387505, LR = 0.005\n", "Epoch 8: loss: 0.06705008948483737, Train Accuracy: 0.9822514379622022, Test Accuracy: 0.9931008049060943, LR = 0.005\n", "Epoch 9: loss: 0.06327662676631007, Train Accuracy: 0.980936729663106, Test Accuracy: 0.9915676504407819, LR = 0.005\n", "Epoch 10: loss: 0.05715959312995741, Train Accuracy: 0.9829087921117502, Test Accuracy: 0.9927175162897662, LR = 0.0025\n", "Epoch 11: loss: 0.04729501416538066, Train Accuracy: 0.9848808545603944, Test Accuracy: 0.9931008049060943, LR = 0.0025\n", "Epoch 12: loss: 0.05846806700719753, Train Accuracy: 0.9853738701725555, Test Accuracy: 0.9923342276734382, LR = 0.0025\n", "Epoch 13: loss: 0.08299519409289739, Train Accuracy: 0.9847165160230074, Test Accuracy: 0.9934840935224224, LR = 0.0025\n", "Epoch 14: loss: 0.05358204972678019, Train Accuracy: 0.9855382087099425, Test Accuracy: 0.9923342276734382, LR = 0.0025\n", "Epoch 15: loss: 0.05235497250396293, Train Accuracy: 0.9842235004108464, Test Accuracy: 0.9938673821387505, LR = 0.0025\n", "Epoch 16: loss: 0.04645653028516487, Train Accuracy: 0.9870172555464256, Test Accuracy: 0.9938673821387505, LR = 0.0025\n", "Epoch 17: loss: 0.04011241256375797, Train Accuracy: 0.9878389482333607, Test Accuracy: 0.9927175162897662, LR = 0.0025\n", "Epoch 18: loss: 0.09714725045826829, Train Accuracy: 0.9857025472473295, Test Accuracy: 0.9942506707550786, LR = 0.0025\n", "Epoch 19: loss: 0.05097170710117401, Train Accuracy: 0.9847165160230074, Test Accuracy: 0.9927175162897662, LR = 0.0025\n", "Epoch 20: loss: 0.06390207394967244, Train Accuracy: 0.9866885784716516, Test Accuracy: 0.9946339593714066, LR = 0.00125\n" ] } ], "source": [ "train(20)" ] }, { "cell_type": "code", "execution_count": 26, "id": "b27ba160", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Model saved to checkpoints/hand_gesture_model.pth\n" ] } ], "source": [ "import os\n", "def save_model(model, filename=\"hand_gesture_model.pth\"):\n", " if not os.path.exists(\"checkpoints\"):\n", " os.makedirs(\"checkpoints\")\n", " \n", " filepath = os.path.join(\"checkpoints\", filename)\n", " torch.save(model.state_dict(), filepath)\n", " print(f\"Model saved to {filepath}\")\n", "\n", "save_model(model)" ] }, { "cell_type": "code", "execution_count": null, "id": "f24e6363", "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "default", "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": "3.12.1" } }, "nbformat": 4, "nbformat_minor": 5 }