debug_predict.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predict Segmentation "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import random\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import keras.backend as K\n",
    "from keras.models import Model\n",
    "\n",
    "from model_fcn import resnet50_fcn, resnet50_16s_fcn, resnet50_8s_fcn\n",
    "from data_generator import seg_data_generator, pad_image\n",
    "import utils\n",
    "\n",
    "model_input_dir = '/work/fcn_models/segmentation_32s'\n",
    "gpu_id = 0\n",
    "\n",
    "img_dir = '/data/COCO/Segmentation/'\n",
    "utils.config_tf()\n",
    "\n",
    "df = pd.read_csv(img_dir + 'labels.txt',header=None)\n",
    "n_classes = len(df.index)\n",
    "\n",
    "net = \"resnet50\"\n",
    "\n",
    "#from IPython.core.debugger import Tracer\n",
    "#Tracer()()\n",
    "# create model\n",
    "#gpu = '/gpu:' + str(gpu_id)\n",
    "#with tf.device(gpu):\n",
    "if net == 'resnet50' : \n",
    "    model, stride = resnet50_fcn(n_classes)\n",
    "\n",
    "if net == 'resnet50_16s' :\n",
    "    model, stride = resnet50_16s_fcn(n_classes)\n",
    "\n",
    "if net == 'resnet50_8s' :\n",
    "    model, stride = resnet50_8s_fcn(n_classes) \n",
    "\n",
    "if model_input_dir != '':\n",
    "    model.load_weights(model_input_dir + '/best_weights.hdf5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "img_id = 0\n",
    "n_rows = 5\n",
    "\n",
    "\n",
    "# load image and predict\n",
    "val_img_dir = img_dir + 'val_images/'\n",
    "val_label_dir = img_dir + 'val_labels/'\n",
    "\n",
    "img_list_val = os.listdir(val_img_dir)\n",
    "img_list_val = img_list_val[img_id:]\n",
    "pred_gen = seg_data_generator(stride,n_classes,val_img_dir,val_label_dir,img_list_val)\n",
    "val_gen = seg_data_generator(stride,n_classes,val_img_dir,val_label_dir,img_list_val,preprocess = False)\n",
    "\n",
    "# Visualise in tiled plot\n",
    "fig = plt.figure(figsize=(15,15))\n",
    "\n",
    "for idx in range(n_rows) :\n",
    "\tfor row_img in range(2) :\n",
    "\t\tx_img, _ = val_gen.next()\n",
    "\t\tx_img = np.array(x_img,dtype='uint8')\n",
    "\n",
    "\t\tx, y = pred_gen.next()\n",
    "\t\ty = np.argmax(y,axis =-1)[0]\n",
    "\n",
    "\t\tpred = model.predict(x,batch_size=1)[0]\n",
    "\t\tpred = np.argmax(pred,axis =-1)\n",
    "\n",
    "\t\t# Visualise result\n",
    "\t\tfig.add_subplot(n_rows,6,idx*6 + row_img*3 +1)\n",
    "\t\tplt.imshow(x_img[0])\n",
    "\t\t\n",
    "\t\tfig.add_subplot(n_rows,6,idx*6 + row_img*3 + 2)\n",
    "\t\tplt.imshow(y, vmin=0,vmax=n_classes-1)\n",
    "\t\t\n",
    "\t\tfig.add_subplot(n_rows,6,idx*6 + row_img*3 + 3)\n",
    "\t\tplt.imshow(pred, vmin=0,vmax=n_classes-1)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predict Pose Estimation "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import random\n",
    "import os\n",
    "\n",
    "import matplotlib as mpl\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import tensorflow as tf\n",
    "import keras.backend as K\n",
    "from keras.models import Model\n",
    "\n",
    "from model_fcn import resnet50_fcn, resnet50_16s_fcn, resnet50_8s_fcn\n",
    "from data_generator import pose_data_generator, pad_image\n",
    "import utils\n",
    "\n",
    "model_input_dir = '/work/fcn_models/pose_estimation_32s'\n",
    "gpu_id = 0\n",
    "\n",
    "n_joints = 14\n",
    "\n",
    "net = \"resnet50\"\n",
    "root_dir = '/home/tdnguyen'\n",
    "root_dir = ''\n",
    "img_dir = root_dir + '/data/lsp_dataset/images/'\n",
    "\n",
    "utils.config_tf()\n",
    "\n",
    "# create model\n",
    "gpu = '/gpu:' + str(gpu_id)\n",
    "#with tf.device(gpu):\n",
    "if net == 'resnet50' : \n",
    "    model, stride = resnet50_fcn(n_joints)\n",
    "\n",
    "if net == 'resnet50_16s' :\n",
    "    model, stride = resnet50_16s_fcn(n_joints)\n",
    "    \n",
    "if net == 'resnet50_8s' :\n",
    "    model, stride = resnet50_8s_fcn(n_joints) \n",
    "\n",
    "if model_input_dir != '':\n",
    "    #debugger.set_trace()\n",
    "    model.load_weights(model_input_dir + '/best_weights.hdf5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "img_id = 15\n",
    "n_rows = 5\n",
    "n_cols = 4\n",
    "\n",
    "# load image and predict\n",
    "N_train_img = 1000\n",
    "img_dir = root_dir + '/data/lsp_dataset/images/'\n",
    "label_file = root_dir + '/data/lsp_dataset/joints.mat'\n",
    "img_list = range(1,2000)\n",
    "\n",
    "pred_gen = pose_data_generator(stride,n_joints,img_dir,label_file,\n",
    "\t\t\t\t\t\t\t  img_list[N_train_img+img_id:])\n",
    "val_gen = pose_data_generator(stride,n_joints,img_dir,label_file,\n",
    "\t\t\t\t\t\t\t  img_list[N_train_img+img_id:],\n",
    "\t\t\t\t\t\t\t  preprocess=False)\n",
    "\n",
    "# Visualise in tiled plot\n",
    "fig = plt.figure(figsize=(15,15))\n",
    "\n",
    "for idx in range(n_rows) :\n",
    "\tfor row_img in range(2) :\n",
    "\t\tx_img, _ = val_gen.next()\n",
    "\t\t\n",
    "\t\tx, y = pred_gen.next()\n",
    "\t\ty = np.argmax(y,axis =-1)[0]\n",
    "\n",
    "\t\tpred = model.predict(x,batch_size=1)[0]\n",
    "\t\t\n",
    "\t\t# Visualise result\n",
    "\t\tfig.add_subplot(n_rows,6,idx*6 + row_img*3 +1)\n",
    "\t\tplt.imshow(x_img[0])\n",
    "\t\t\n",
    "\t\tfig.add_subplot(n_rows,6,idx*6 + row_img*3 + 2)\n",
    "\t\tplt.imshow(y, vmin=0,vmax=n_joints-1)\n",
    "\t\t\t\n",
    "\t\tfig.add_subplot(n_rows,6,idx*6 + row_img*3 + 3)\n",
    "\t\tplt.imshow(pred[:,:,0], vmin=0,vmax=1)\n",
    "\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}