{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# Clean Human Pose MPI Dataset\n", "\n", "Fix data\n", "\n", "Data structure:\n", "- `data[2]` = 2 x 7 x 100 array\n", "- `data[2][0]` = x locations\n", "- `data[2][0]` = y locations\n", "- ordering is `0 Head, 1 Right wrist, 2 Left wrist, 3 Right elbow, 4 Left elbow, 5 Right shoulder and 6 Left shoulder`" ] }, { "cell_type": "code", "execution_count": 175, "metadata": {}, "outputs": [], "source": [ "%reload_ext autoreload\n", "%autoreload 2\n", "\n", "import os\n", "from os.path import join\n", "import math\n", "from glob import glob\n", "from random import randint\n", "\n", "import cv2 as cv\n", "import numpy as np\n", "import pandas as pd\n", "from PIL import Image, ImageDraw\n", "%matplotlib inline\n", "import matplotlib.pyplot as plt\n", "import scipy.io\n", "from pathlib import Path\n", "from sklearn import preprocessing\n", "\n", "import sys\n", "sys.path.append('/work/megapixels_dev/megapixels/')\n", "from app.settings import app_cfg as cfg\n", "from app.utils import file_utils" ] }, { "cell_type": "code", "execution_count": 176, "metadata": {}, "outputs": [], "source": [ "DATA_STORE = '/data_store_nas/'\n", "fp_dataset = join(DATA_STORE, 'datasets/people/youtube_poses')\n", "dir_fp_frames = join(fp_dataset, 'YouTube_Pose_dataset_1.0/GT_frames')" ] }, { "cell_type": "code", "execution_count": 177, "metadata": {}, "outputs": [], "source": [ "dirs_frames = glob(join(dir_fp_frames, '*'))\n", "fps_frames = {}\n", "for dir_frames in dirs_frames:\n", " fps_frames[dir_frames] = join(dir_frames, '*')" ] }, { "cell_type": "code", "execution_count": 178, "metadata": {}, "outputs": [], "source": [ "fp_pose_data = join(fp_dataset, 'YouTube_Pose_dataset_1.0/YouTube_Pose_dataset.mat')\n", "fp_out = join(fp_dataset, 'poses.csv')\n", "pose_data = scipy.io.loadmat(fp_pose_data)['data'][0]" ] }, { "cell_type": "code", "execution_count": 182, "metadata": {}, "outputs": [], "source": [ "# convert data to pandas DF for sanity\n", "poses = []\n", "for i, pose in enumerate(pose_data):\n", "\n", " video_id = pose[1][0]\n", " pose_pts = pose[2]\n", " crop_x1 = int(pose[6][0][0])\n", " crop_y1 = int(pose[6][0][1])\n", " crop_x2 = int(pose[6][0][2])\n", " crop_y2 = int(pose[6][0][3])\n", " w = pose[7][0][0]\n", " h = pose[7][0][1]\n", " scale = pose[5][0][0]\n", " \n", " for j in range(pose_pts.shape[2]): # 100 frames\n", " x = [pose_pts[0][i][j] for i in range(7)]\n", " y = [pose_pts[1][i][j] for i in range(7)]\n", " poses.append({\n", " 'video_id': video_id, \n", " 'scale': scale,\n", " 'crop_x1': crop_x1,\n", " 'crop_y1': crop_y1,\n", " 'crop_x2': crop_x2,\n", " 'crop_y2': crop_y2,\n", " 'width': w, \n", " 'height': h,\n", " 'head_x': x[0],\n", " 'head_y': y[0],\n", " 'wrist_right_x': x[1],\n", " 'wrist_right_y': y[1],\n", " 'wrist_left_x': x[2], \n", " 'wrist_left_y': y[2],\n", " 'elbow_right_x': x[3],\n", " 'elbow_right_y': y[3],\n", " 'elbow_left_x': x[4], \n", " 'elbow_left_y': y[4],\n", " 'shoulder_right_x': x[5],\n", " 'shoulder_right_y': y[5],\n", " 'shoulder_left_x': x[6], \n", " 'shoulder_left_y': y[6],\n", " })\n", "df_poses = pd.DataFrame.from_dict(poses)\n", "df_poses.to_csv(fp_out)" ] }, { "cell_type": "code", "execution_count": 183, "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", "
crop_x1crop_x2crop_y1crop_y2elbow_left_xelbow_left_yelbow_right_xelbow_right_yhead_xhead_y...shoulder_left_xshoulder_left_yshoulder_right_xshoulder_right_yvideo_idwidthwrist_left_xwrist_left_ywrist_right_xwrist_right_y
01192011080277.721438192.416331147.628696169.326277195.49832081.471438...254.631384127.088374178.603159134.691196-osma2n86oA720278.566196235.498992158.047379122.301411
11192011080273.497648187.629368152.134073129.341062207.32493372.742272...254.349798131.593750181.137433123.990927-osma2n86oA720274.342406235.498992135.23891191.608535
21192011080258.010417159.752352160.581653143.138777229.00705676.966062...250.407594125.117272190.992944117.232863-osma2n86oA720213.801411108.785282181.98219189.074261
31192011080274.342406188.192540142.841734110.193212203.10114276.402890...253.786626128.777890185.361223120.611895-osma2n86oA720276.595094231.556788156.92103555.847110
41192011080272.371304194.387433225.628024164.820901245.90221893.016465...255.476142139.478159183.390121126.806788-osma2n86oA720305.316868172.423723278.284610165.102487
\n", "

5 rows × 22 columns

\n", "
" ], "text/plain": [ " crop_x1 crop_x2 crop_y1 crop_y2 elbow_left_x elbow_left_y \\\n", "0 1 1920 1 1080 277.721438 192.416331 \n", "1 1 1920 1 1080 273.497648 187.629368 \n", "2 1 1920 1 1080 258.010417 159.752352 \n", "3 1 1920 1 1080 274.342406 188.192540 \n", "4 1 1920 1 1080 272.371304 194.387433 \n", "\n", " elbow_right_x elbow_right_y head_x head_y ... \\\n", "0 147.628696 169.326277 195.498320 81.471438 ... \n", "1 152.134073 129.341062 207.324933 72.742272 ... \n", "2 160.581653 143.138777 229.007056 76.966062 ... \n", "3 142.841734 110.193212 203.101142 76.402890 ... \n", "4 225.628024 164.820901 245.902218 93.016465 ... \n", "\n", " shoulder_left_x shoulder_left_y shoulder_right_x shoulder_right_y \\\n", "0 254.631384 127.088374 178.603159 134.691196 \n", "1 254.349798 131.593750 181.137433 123.990927 \n", "2 250.407594 125.117272 190.992944 117.232863 \n", "3 253.786626 128.777890 185.361223 120.611895 \n", "4 255.476142 139.478159 183.390121 126.806788 \n", "\n", " video_id width wrist_left_x wrist_left_y wrist_right_x wrist_right_y \n", "0 -osma2n86oA 720 278.566196 235.498992 158.047379 122.301411 \n", "1 -osma2n86oA 720 274.342406 235.498992 135.238911 91.608535 \n", "2 -osma2n86oA 720 213.801411 108.785282 181.982191 89.074261 \n", "3 -osma2n86oA 720 276.595094 231.556788 156.921035 55.847110 \n", "4 -osma2n86oA 720 305.316868 172.423723 278.284610 165.102487 \n", "\n", "[5 rows x 22 columns]" ] }, "execution_count": 183, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df_poses.head()" ] }, { "cell_type": "code", "execution_count": 172, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "22" ] }, "execution_count": 172, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(df_poses.keys())" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "megapixels", "language": "python", "name": "megapixels" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.7.1" } }, "nbformat": 4, "nbformat_minor": 2 }