(option)local_predict.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# モデルの URI とコンテナイメージの取得\n",
    "import boto3\n",
    "sm_client = boto3.client('sagemaker')\n",
    "endpoint_config_name = sm_client.describe_endpoint(EndpointName='jumpstart-ftc-kinoko-detection-endpoint')['EndpointConfigName']\n",
    "model_name = sm_client.describe_endpoint_config(EndpointConfigName=endpoint_config_name)['ProductionVariants'][0]['ModelName']\n",
    "model_detail = sm_client.describe_model(ModelName=model_name)\n",
    "model_container_image = model_detail['PrimaryContainer']['Image']\n",
    "model_s3_uri = model_detail['PrimaryContainer']['ModelDataUrl']\n",
    "print(model_container_image)\n",
    "print(model_s3_uri)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# モデルのダウンロード\n",
    "!rm -rf model/\n",
    "!mkdir -p model/\n",
    "!aws s3 cp {model_s3_uri} ./model/\n",
    "%cd model/\n",
    "!tar zxvf model.tar.gz"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# 推論準備\n",
    "%cd code\n",
    "!pip install sagemaker-inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 依存しているライブラリをインストール\n",
    "!pip install sagemaker-inference\n",
    "\n",
    "import json, inference\n",
    "from PIL import Image, ImageDraw, ImageFont\n",
    "\n",
    "# モデルのロード\n",
    "model = inference.model_fn('../')\n",
    "# 推論に使う画像ファイル\n",
    "IMAGE_FILE = '../../test_raw_images/lattice.jpg'\n",
    "# 画像ファイルをバイナリとして開く\n",
    "with open(IMAGE_FILE,'rb') as f:\n",
    "    img_bin = f.read()\n",
    "# 推論\n",
    "pred = json.loads(inference.transform_fn(model,img_bin,'application/x-image','application/json'))\n",
    "# 推論結果の展開\n",
    "bboxes, classes, confidences = pred['normalized_boxes'], pred['classes'], pred['scores']\n",
    "# 推論結果の描画\n",
    "img = Image.open(IMAGE_FILE)\n",
    "draw = ImageDraw.Draw(img)\n",
    "for i in range(len(bboxes)):\n",
    "    # 信頼度スコアが0.8以上のみ描画する\n",
    "    if confidences[i]>0.8:\n",
    "        # 矩形の相対座標を取得\n",
    "        left, top, right, bottom = bboxes[i]\n",
    "        # 矩形の相対座標を絶対座標に変換\n",
    "        left = img.size[0] * left\n",
    "        top = img.size[1] * top\n",
    "        right = img.size[0] * right\n",
    "        bottom = img.size[1] * bottom\n",
    "        # 検出した物体の ID を take/kino に読み替える\n",
    "        text = 'take' if int(classes[i])==0 else 'kino'\n",
    "        # take/kinoに対して矩形で描画するための色を設定する\n",
    "        color = 'blue' if text == 'take' else 'red'\n",
    "        # 矩形の左上に表示する文字の大きさを設定、きのこの山なら大きくする\n",
    "        TEXTSIZE=14 if classes[i]=='0' else 18\n",
    "        # 矩形の先の太さを設定、きのこの山なら太くする\n",
    "        LINEWIDTH=4 if classes[i]=='0' else 6\n",
    "        # 矩形を描画する\n",
    "        draw.rectangle([(left,top),(right,bottom)], outline=color, width=LINEWIDTH)\n",
    "        # 矩形の左上に描画する信頼度スコアの取得\n",
    "        text += f' {str(round(confidences[i],3))}'\n",
    "        # テキストを描画する場所を取得\n",
    "        txpos = (left, top-TEXTSIZE-LINEWIDTH//2)\n",
    "        # フォントの設定\n",
    "        font = ImageFont.truetype(\"/usr/share/fonts/truetype/dejavu/DejaVuSerif.ttf\", size=TEXTSIZE)\n",
    "        # 描画するテキストのサイズを取得\n",
    "        txw, txh = draw.textsize(text, font=font)\n",
    "        # テキストの背景用の矩形を描画\n",
    "        draw.rectangle([txpos, (left+txw, top)], outline=color, fill=color, width=LINEWIDTH)\n",
    "        # テキストを描画\n",
    "        draw.text(txpos, text, fill='white',font=font)\n",
    "# 画像をファイルに書き出す\n",
    "img.save('pred_lattice.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "img"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "instance_type": "ml.t3.medium",
  "kernelspec": {
   "display_name": "Python 3 (MXNet 1.8 Python 3.7 CPU Optimized)",
   "language": "python",
   "name": "python3__SAGEMAKER_INTERNAL__arn:aws:sagemaker:us-east-2:429704687514:image/mxnet-1.8-cpu-py37-ubuntu16.04-v1"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}