test_tflite_models.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[{'name': 'input_1', 'index': 0, 'shape': array([  1, 224, 224,   3], dtype=int32), 'shape_signature': array([  1, 224, 224,   3], dtype=int32), 'dtype': <class 'numpy.uint8'>, 'quantization': (0.0, 0), 'quantization_parameters': {'scales': array([], dtype=float32), 'zero_points': array([], dtype=int32), 'quantized_dimension': 0}, 'sparsity_parameters': {}}]\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import matplotlib.pylab as plt\n",
    "import cv2\n",
    "import numpy as np\n",
    "\n",
    "tflite_model_file = 'openpose_singlenet.tflite'\n",
    "\n",
    "interpreter = tf.lite.Interpreter(model_path=tflite_model_file)\n",
    "\n",
    "inp_index = interpreter.get_input_details()[0][\"index\"]\n",
    "\n",
    "out_details = sorted(interpreter.get_output_details(), key=lambda k: k['index']) \n",
    "\n",
    "heatmap_idx = out_details[-1][\"index\"]\n",
    "paf_idx = out_details[-2][\"index\"]\n",
    "\n",
    "print(interpreter.get_input_details())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load sample image and run the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_image = 'resources/ski_224.jpg'\n",
    "\n",
    "img = cv2.imread(test_image) # B,G,R order\n",
    "img = np.expand_dims(img, 0)\n",
    "\n",
    "input_tensor= tf.convert_to_tensor(img, np.uint8)\n",
    "\n",
    "interpreter.allocate_tensors()\n",
    "\n",
    "interpreter.set_tensor(inp_index, input_tensor)\n",
    "\n",
    "interpreter.invoke()\n",
    "\n",
    "heatmaps = interpreter.get_tensor(heatmap_idx)\n",
    "pafs = interpreter.get_tensor(paf_idx)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Show predictions: paf and heatmap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f1b044a7650>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "heatmap_idx = 0 # nose\n",
    "\n",
    "plt.imshow(heatmaps[0, :, :, heatmap_idx], cmap='gray')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f1b65280890>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "paf_dx_idx = 0\n",
    "paf_dy_idx = 1\n",
    "\n",
    "plt.imshow(pafs[0, :, :, paf_dx_idx], cmap='gray')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f1b652677d0>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(pafs[0, :, :, paf_dy_idx], cmap='gray')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}