{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Demo"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Set Up\n",
    "prerequisites\n",
    "- python 3.12\n",
    "\n",
    "``` shell\n",
    "$ cd ./package/{platform}/\n",
    "$ pip install KneronPLUS-2.3.0-py3-none-any.whl\n",
    "$ pip install --force-reinstall KneronPLUS-2.3.0-py3-none-any.whl\n",
    "$ pip install opencv-pythn\n",
    "\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Connect KL520"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2\n",
    "import kp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22\n"
     ]
    }
   ],
   "source": [
    "device_descriptors = kp.core.scan_devices()\n",
    "\n",
    "if 0 < device_descriptors.device_descriptor_number:\n",
    "    usb_port_id = device_descriptors.device_descriptor_list[0].usb_port_id\n",
    "    print(usb_port_id)\n",
    "else:\n",
    "    print('Error: no Kneron device connect.')\n",
    "    exit(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "device_group = kp.core.connect_devices(usb_port_ids=[22])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "kp.core.set_timeout(device_group=device_group,\n",
    "                    milliseconds=5000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load firmware, model, and test image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "SCPU_FW_PATH = '../../res/firmware/KL520/fw_scpu.bin'\n",
    "NCPU_FW_PATH = '../../res/firmware/KL520/fw_ncpu.bin'\n",
    "kp.core.load_firmware_from_file(device_group=device_group,\n",
    "                                scpu_fw_path=SCPU_FW_PATH,\n",
    "                                ncpu_fw_path=NCPU_FW_PATH)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "MODEL_FILE_PATH = './photo_scorer_520.nef'\n",
    "model_nef_descriptor = kp.core.load_model_from_file(device_group=device_group,\n",
    "                                                    file_path=MODEL_FILE_PATH)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### pre-processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "IMAGE_FILE_PATH = './test.jpg'\n",
    "\n",
    "img = cv2.imread(filename=IMAGE_FILE_PATH)\n",
    "img_bgr565 = cv2.cvtColor(src=img, code=cv2.COLOR_BGR2BGR565)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "generic_inference_input_descriptor = kp.GenericImageInferenceDescriptor(\n",
    "    model_id=model_nef_descriptor.models[0].id,\n",
    "    inference_number=0,\n",
    "    input_node_image_list=[\n",
    "        kp.GenericInputNodeImage(\n",
    "            image=img_bgr565,\n",
    "            image_format=kp.ImageFormat.KP_IMAGE_FORMAT_RGB565,\n",
    "            resize_mode=kp.ResizeMode.KP_RESIZE_ENABLE,\n",
    "            padding_mode=kp.PaddingMode.KP_PADDING_CORNER,\n",
    "            normalize_mode=kp.NormalizeMode.KP_NORMALIZE_KNERON\n",
    "        )\n",
    "    ]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "    \"header\": {\n",
      "        \"inference_number\": 0,\n",
      "        \"crop_number\": 0,\n",
      "        \"num_output_node\": 1,\n",
      "        \"product_id\": 256,\n",
      "        \"num_hw_pre_proc_info\": 1,\n",
      "        \"hw_pre_proc_info_list\": {\n",
      "            \"0\": {\n",
      "                \"img_width\": 1200,\n",
      "                \"img_height\": 800,\n",
      "                \"resized_img_width\": 224,\n",
      "                \"resized_img_height\": 149,\n",
      "                \"pad_top\": 0,\n",
      "                \"pad_bottom\": 75,\n",
      "                \"pad_left\": 0,\n",
      "                \"pad_right\": 0,\n",
      "                \"model_input_width\": 224,\n",
      "                \"model_input_height\": 224,\n",
      "                \"crop_area\": {\n",
      "                    \"crop_box_index\": 0,\n",
      "                    \"x\": 0,\n",
      "                    \"y\": 0,\n",
      "                    \"width\": 0,\n",
      "                    \"height\": 0\n",
      "                }\n",
      "            }\n",
      "        }\n",
      "    },\n",
      "    \"raw_result\": {\n",
      "        \"buffer_size\": 388\n",
      "    }\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "kp.inference.generic_image_inference_send(device_group=device_group,\n",
    "                                          generic_inference_input_descriptor=generic_inference_input_descriptor)\n",
    "generic_raw_result = kp.inference.generic_image_inference_receive(device_group=device_group)\n",
    "print(generic_raw_result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[{\n",
      "    \"width\": 1,\n",
      "    \"height\": 1,\n",
      "    \"channel\": 1,\n",
      "    \"channels_ordering\": \"ChannelOrdering.KP_CHANNEL_ORDERING_CHW\",\n",
      "    \"num_data\": 1,\n",
      "    \"ndarray\": [\n",
      "        \"[[[[0.71443015]]]]\"\n",
      "    ]\n",
      "}]\n"
     ]
    }
   ],
   "source": [
    "inf_node_output_list = []\n",
    "\n",
    "for node_idx in range(generic_raw_result.header.num_output_node):\n",
    "    inference_float_node_output = kp.inference.generic_inference_retrieve_float_node(node_idx=node_idx,\n",
    "                                                                                     generic_raw_result=generic_raw_result,\n",
    "                                                                                     channels_ordering=kp.ChannelOrdering.KP_CHANNEL_ORDERING_CHW)\n",
    "    inf_node_output_list.append(inference_float_node_output)\n",
    "\n",
    "print(inf_node_output_list)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### post-processing\n",
    "Kneron PLUS python version doesn't support on-NPU post-processing, so this step is demonstrated here"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The score of this photo is 0.7144301533699036\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "\n",
    "nd_array = inf_node_output_list[0].ndarray\n",
    "number = float(nd_array.flatten()[0])\n",
    "\n",
    "print(\"The score of this photo is\", number)"
   ]
  }
 ],
 "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": "3.11.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}