cifar_caffe_nvdla_feature_maps_compare_42_images.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import sys\n",
    "import os\n",
    "import caffe\n",
    "import time\n",
    "import glob\n",
    "import scipy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def top_k(result, top_num, label_arr):\n",
    "    tmp = -(top_num+1)\n",
    "    top_k = result.flatten().argsort()[-1:tmp:-1]\n",
    "    top_num = 1\n",
    "    for i in np.arange(top_k.size):\n",
    "        print(\"Top %d:\\tProb:%2.4f%%\\t[%d]\\t %s\"%(top_num,result[top_k[i]]*100,top_k[i], labels[top_k[i]]))\n",
    "        top_num = top_num + 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_files(path,fileType):\n",
    "    dimg_files = []\n",
    "    for root, directories, filenames in os.walk(path):\n",
    "        for filename in filenames:\n",
    "            if fileType in filename:\n",
    "                dimg_files.append(os.path.join(root,filename))\n",
    "    return dimg_files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_dimg(dimg_file):\n",
    "    return np.genfromtxt(dimg_file, float, delimiter=' ')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_image_path(basename,path,ext):\n",
    "    return(path + os.path.basename(basename).split('.')[0] + ext)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def loss_rate(arr1, arr2):\n",
    "#     print arr1, arr2\n",
    "    return np.sum(np.absolute(np.subtract(arr1,arr2)))/len(arr1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def cross_entropy(predictions, targets, epsilon=1e-12):\n",
    "    \"\"\"\n",
    "    Computes cross entropy between targets (encoded as one-hot vectors)\n",
    "    and predictions. \n",
    "    Input: predictions (N, k) ndarray\n",
    "           targets (N, k) ndarray        \n",
    "    Returns: scalar\n",
    "    \"\"\"\n",
    "    predictions = np.clip(predictions, epsilon, 1. - epsilon)\n",
    "    N = predictions.shape[0]\n",
    "    ce = -np.sum(np.sum(targets*np.log(predictions+1e-9)))/N\n",
    "    return ce"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# function is to show layer data by gray image\n",
    "def show_data(data, padsize=1, padval=0):\n",
    "    data -= data.min()\n",
    "    data /= data.max()\n",
    "    \n",
    "    # force the number of filters to be square\n",
    "    n = int(np.ceil(np.sqrt(data.shape[0])))\n",
    "    padding = ((0, n ** 2 - data.shape[0]), (0, padsize), (0, padsize)) + ((0, 0),) * (data.ndim - 3)\n",
    "    data = np.pad(data, padding, mode='constant', constant_values=(padval, padval))\n",
    "    \n",
    "    # tile the filters into an image\n",
    "    data = data.reshape((n, n) + data.shape[1:]).transpose((0, 2, 1, 3) + tuple(range(4, data.ndim + 1)))\n",
    "    data = data.reshape((n * data.shape[1], n * data.shape[3]) + data.shape[4:])\n",
    "    plt.figure()\n",
    "    plt.imshow(data,cmap='gray')\n",
    "    plt.axis('on')\n",
    "plt.rcParams['figure.figsize'] = (8, 8)\n",
    "plt.rcParams['image.interpolation'] = 'nearest'\n",
    "plt.rcParams['image.cmap'] = 'gray'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def caffe_inference(image_path):\n",
    "    \n",
    "    # Configure\n",
    "    pycaffe_dir = '/home/caslab/workspace/caffe/caffe/python/'\n",
    "    model_dir = '/home/caslab/workspace/NVDLA/sw_5_17/sw/prebuilt/linux/model/'\n",
    "    input_file = image_path\n",
    "    model_def = os.path.join(model_dir,\"cifar10/cifar_quick_conv.prototxt\")\n",
    "    pretrained_model = os.path.join(model_dir,\"cifar10/cifar10_quick_iter_4000.caffemodel\")\n",
    "    gpu = False\n",
    "    center_only = True\n",
    "    images_dim = '32,32'\n",
    "    mean_file = '/home/caslab/workspace/caffe/caffe/examples/cifar10/mean.npy'\n",
    "    # input_scale = \n",
    "    raw_scale = 255.0\n",
    "    channel_swap = '2,1,0'\n",
    "    ext = 'jpg'\n",
    "    image_dims = [int(s) for s in images_dim.split(',')]\n",
    "    mean, channel_swap_arr = None, None\n",
    "    \n",
    "    \n",
    "    \n",
    "    # Setup parameter\n",
    "    if mean_file:\n",
    "        mean = np.load(mean_file)\n",
    "    if channel_swap:\n",
    "        channel_swap_arr = [int(s) for s in channel_swap.split(',')]\n",
    "    if gpu:\n",
    "        caffe.set_mode_gpu()\n",
    "    else:\n",
    "        caffe.set_mode_cpu()\n",
    "\n",
    "    \n",
    "    # Create classifier class\n",
    "    classifier = caffe.Classifier(model_def, pretrained_model,\n",
    "            image_dims=image_dims, mean=None,\n",
    "    #         image_dims=image_dims, mean=mean,\n",
    "            raw_scale=raw_scale,\n",
    "            channel_swap=channel_swap_arr)\n",
    "\n",
    "\n",
    "    \n",
    "    # Load image file.\n",
    "    inputs =[caffe.io.load_image(input_file)]\n",
    "    load_img_time_end = time.time()\n",
    "\n",
    "    # Classify\n",
    "    predictions = classifier.predict(inputs, not center_only)\n",
    "    \n",
    "    return predictions[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "dimg_root_path = '/home/caslab/workspace/NVDLA/sw_6_1/prebuilt/linux/runtime/test/'\n",
    "\n",
    "image_root_path = '/home/caslab/workspace/dataset/cifar/mix/'\n",
    "\n",
    "test_name = []\n",
    "nvdla_result = []\n",
    "caffe_result = []\n",
    "\n",
    "for element in dimg_files:\n",
    "    test_name.append(os.path.basename(element).split('.')[0])\n",
    "    nvdla_result.append(read_dimg(element))\n",
    "    caffe_result.append(caffe_inference(get_image_path(element,image_root_path,'.jpg')))\n",
    "#     caffe_nvdla_loss.append(loss_rate(read_dimg(element).reshape(-1),caffe_inference(get_image_path(element,image_root_path,'.jpg')).reshape(-1)))\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# caffe_nvdla_loss = []\n",
    "# for z in range(0,len(nvdla_result)):\n",
    "#     caffe_nvdla_loss.append(loss_rate(nvdla_result[z],caffe_result[z].reshape(-1)))\n",
    "\n",
    "caffe_nvdla_loss = []\n",
    "\n",
    "for z in range(0,len(nvdla_result)):\n",
    "    caffe_nvdla_loss.append(loss_rate(nvdla_result[z],caffe_result[z].reshape(-1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa5cd6f5390>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pixel_width = 6\n",
    "pixel_height = 6\n",
    "plt.figure(figsize=(pixel_width, pixel_height))\n",
    "aa = range(0,42)\n",
    "l1, = plt.plot(aa,caffe_nvdla_loss,'ro', label = 'Loss rate', color = 'blue')\n",
    "\n",
    "plt.xlim((0, 41))\n",
    "plt.ylim((0, 1))\n",
    "plt.title(\"Absolute Loss between caffe and nvdla with 42 images\")\n",
    "plt.xlabel('images')\n",
    "plt.ylabel('Absolute Loss')\n",
    "\n",
    "plt.legend() #, labels = ['caffe', 'no mean'], loc = 'best')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}