GluonCV Human Pose.ipynb

gluoncv-human-pose.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "GluonCV Human Pose.ipynb",
      "provenance": [],
      "private_outputs": true,
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyPBkf+YrdikDI3lgVd0yk+r",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/bedilbek/3937154a784375471e935dc830abf489/gluoncv-human-pose.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5VVMsTJ7n1PA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!pip install --upgrade mxnet gluoncv"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vxeVNi_jn-zR",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from matplotlib import pyplot as plt\n",
        "from gluoncv import model_zoo, data, utils\n",
        "from gluoncv.data.transforms.pose import detector_to_alpha_pose, heatmap_to_coord_alpha_pose"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "emN8mtG8pNp1",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "detector = model_zoo.get_model('yolo3_mobilenet1.0_coco', pretrained=True)\n",
        "pose_net = model_zoo.get_model('alpha_pose_resnet101_v1b_coco', pretrained=True)\n",
        "\n",
        "# Note that we can reset the classes of the detector to only include\n",
        "# human, so that the NMS process is faster.\n",
        "\n",
        "detector.reset_class([\"person\"], reuse_weights=['person'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J3UcSuftpW4g",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "im_fname = utils.download('https://github.com/dmlc/web-data/blob/master/' +\n",
        "                          'gluoncv/pose/soccer.png?raw=true',\n",
        "                          path='soccer.png')\n",
        "x, img = data.transforms.presets.yolo.load_test(im_fname, short=512)\n",
        "print('Shape of pre-processed image:', x.shape)\n",
        "\n",
        "class_IDs, scores, bounding_boxs = detector(x)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iNHEB1EUpZ1F",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "pose_input, upscale_bbox = detector_to_alpha_pose(img, class_IDs, scores, bounding_boxs)\n",
        "predicted_heatmap = pose_net(pose_input)\n",
        "pred_coords, confidence = heatmap_to_coord_alpha_pose(predicted_heatmap, upscale_bbox)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GagPXYvBphzF",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "ax = utils.viz.plot_keypoints(img, pred_coords, confidence,\n",
        "                              class_IDs, bounding_boxs, scores,\n",
        "                              box_thresh=0.5, keypoint_thresh=0.2)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}