detecting_demo.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cNz-yh2HFz_G"
      },
      "source": [
        "GitHub  \n",
        "https://paperswithcode.com/paper/detecting-twenty-thousand-classes-using-image  \n",
        "論文  \n",
        "https://arxiv.org/abs/2201.02605v2  \n",
        "\n",
        "<a href=\"https://colab.research.google.com/github/kaz12tech/ai_demos/blob/master/detecting_demo.ipynb\" target=\"_blank\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mfPfm0f4FzgM",
        "outputId": "2e36f486-aada-47b5-d646-17c29eb0b8c8"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "TORCH_VERSION = \".\".join(torch.__version__.split(\".\")[:2])\n",
        "CUDA_VERSION = torch.__version__.split(\"+\")[-1]\n",
        "print(\"torch: \", TORCH_VERSION, \"; cuda: \", CUDA_VERSION)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "m4o2raKjGeo4"
      },
      "source": [
        "detectron2をインストール"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "lqdZSUZjGd0s",
        "outputId": "9b0692f0-01c6-4993-bc78-7d02512a5f5e"
      },
      "outputs": [],
      "source": [
        "!pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/$CUDA_VERSION/torch$TORCH_VERSION/index.html"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "V14jcibzG4mK"
      },
      "source": [
        "GitHubからコードをclone"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "TZOSKqCVG223",
        "outputId": "d1cc3843-ffe1-47e0-8f77-3aec546b8ffc"
      },
      "outputs": [],
      "source": [
        "# clone and install Detic\n",
        "!git clone https://github.com/facebookresearch/Detic.git --recurse-submodules\n",
        "%cd Detic\n",
        "!pip install -r requirements.txt"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "H439Q2e3r4QE"
      },
      "source": [
        "セットアップ"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "kV_3pb2UHmC4"
      },
      "outputs": [],
      "source": [
        "# Some basic setup:\n",
        "# Setup detectron2 logger\n",
        "import detectron2\n",
        "from detectron2.utils.logger import setup_logger\n",
        "setup_logger()\n",
        "\n",
        "# import some common libraries\n",
        "import sys\n",
        "import numpy as np\n",
        "import os, json, cv2, random\n",
        "from google.colab.patches import cv2_imshow\n",
        "\n",
        "# import some common detectron2 utilities\n",
        "from detectron2 import model_zoo\n",
        "from detectron2.engine import DefaultPredictor\n",
        "from detectron2.config import get_cfg\n",
        "from detectron2.utils.visualizer import Visualizer\n",
        "from detectron2.data import MetadataCatalog, DatasetCatalog\n",
        "\n",
        "# Detic libraries\n",
        "sys.path.insert(0, 'third_party/CenterNet2/projects/CenterNet2/')\n",
        "from centernet.config import add_centernet_config\n",
        "from detic.config import add_detic_config\n",
        "from detic.modeling.utils import reset_cls_test"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sv17jGxDr4QH"
      },
      "source": [
        "検出器の定義と学習済みモデルのダウンロード"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "dbtgXuNFHppz",
        "outputId": "c219b7b4-ad52-444f-f7d4-11b019b976bf"
      },
      "outputs": [],
      "source": [
        "# Build the detector and download our pretrained weights\n",
        "cfg = get_cfg()\n",
        "add_centernet_config(cfg)\n",
        "add_detic_config(cfg)\n",
        "cfg.merge_from_file(\"configs/Detic_LCOCOI21k_CLIP_SwinB_896b32_4x_ft4x_max-size.yaml\")\n",
        "cfg.MODEL.WEIGHTS = 'https://dl.fbaipublicfiles.com/detic/Detic_LCOCOI21k_CLIP_SwinB_896b32_4x_ft4x_max-size.pth'\n",
        "cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5  # set threshold for this model\n",
        "cfg.MODEL.ROI_BOX_HEAD.ZEROSHOT_WEIGHT_PATH = 'rand'\n",
        "cfg.MODEL.ROI_HEADS.ONE_CLASS_PER_PROPOSAL = True # For better visualization purpose. Set to False for all classes.\n",
        "predictor = DefaultPredictor(cfg)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EcAu2AXkr4QL"
      },
      "source": [
        "検出対象のファイルをアップロード"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 203,
          "resources": {
            "http://localhost:8080/nbextensions/google.colab/files.js": {
              "data": "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",
              "headers": [
                [
                  "content-type",
                  "application/javascript"
                ]
              ],
              "ok": true,
              "status": 200,
              "status_text": ""
            }
          }
        },
        "id": "b2G12FE3MbKt",
        "outputId": "97fbad68-dd35-4ad5-b070-fe7ebd51f8ad"
      },
      "outputs": [],
      "source": [
        "from google.colab import files\n",
        "uploaded = files.upload()\n",
        "uploaded = list(uploaded.keys())\n",
        "print(uploaded)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "Afh8uWQFO7d_",
        "outputId": "85c97190-e629-48cb-d878-1647d3a86587"
      },
      "outputs": [],
      "source": [
        "from detic.modeling.text.text_encoder import build_text_encoder\n",
        "def get_clip_embeddings(vocabulary, prompt='a '):\n",
        "    text_encoder = build_text_encoder(pretrain=True)\n",
        "    text_encoder.eval()\n",
        "    texts = [prompt + x for x in vocabulary]\n",
        "    emb = text_encoder(texts).detach().permute(1, 0).contiguous().cpu()\n",
        "    return emb\n",
        "\n",
        "vocabulary = 'custom'\n",
        "metadata = MetadataCatalog.get(\"__unused\")\n",
        "\n",
        "#@title 検出対象の入力\n",
        "#@markdown 検出対象の名称を英語で入力してください。\n",
        "detect_target = 'laptop' #@param {type:\"string\"}\n",
        "\n",
        "metadata.thing_classes = [detect_target]\n",
        "\n",
        "classifier = get_clip_embeddings(metadata.thing_classes)\n",
        "num_classes = len(metadata.thing_classes)\n",
        "reset_cls_test(predictor.model, classifier, num_classes)\n",
        "\n",
        "for file in uploaded:\n",
        "  im = cv2.imread(file)\n",
        "\n",
        "  # Reset visualization threshold\n",
        "  output_score_threshold = 0.3\n",
        "  for cascade_stages in range(len(predictor.model.roi_heads.box_predictor)):\n",
        "    predictor.model.roi_heads.box_predictor[cascade_stages].test_score_thresh = output_score_threshold\n",
        "\n",
        "  # Run model and show results\n",
        "  outputs = predictor(im)\n",
        "  v = Visualizer(im[:, :, ::-1], metadata)\n",
        "  out = v.draw_instance_predictions(outputs[\"instances\"].to(\"cpu\"))\n",
        "  cv2_imshow(out.get_image()[:, :, ::-1])\n",
        "\n",
        "del metadata.thing_classes"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MjzgDXDssY_g"
      },
      "source": [
        "ボキャブラリ(検出用のラベル)を設定"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "q2N0lmrpH_q-",
        "outputId": "d137d568-359f-41ed-ff95-6b370b2f5f3c"
      },
      "outputs": [],
      "source": [
        "# Setup the model's vocabulary using build-in datasets\n",
        "\n",
        "BUILDIN_CLASSIFIER = {\n",
        "    'lvis': 'datasets/metadata/lvis_v1_clip_a+cname.npy',\n",
        "    'objects365': 'datasets/metadata/o365_clip_a+cnamefix.npy',\n",
        "    'openimages': 'datasets/metadata/oid_clip_a+cname.npy',\n",
        "    'coco': 'datasets/metadata/coco_clip_a+cname.npy',\n",
        "}\n",
        "\n",
        "BUILDIN_METADATA_PATH = {\n",
        "    'lvis': 'lvis_v1_val',\n",
        "    'objects365': 'objects365_v2_val',\n",
        "    'openimages': 'oid_val_expanded',\n",
        "    'coco': 'coco_2017_val',\n",
        "}\n",
        "\n",
        "vocabulary = 'lvis' # change to 'lvis', 'objects365', 'openimages', or 'coco'\n",
        "metadata = MetadataCatalog.get(BUILDIN_METADATA_PATH[vocabulary])\n",
        "classifier = BUILDIN_CLASSIFIER[vocabulary]\n",
        "num_classes = len(metadata.thing_classes)\n",
        "reset_cls_test(predictor.model, classifier, num_classes)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OLa5cchxsm1R"
      },
      "source": [
        "ボキャブラリに設定した物体を全て検出"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "xogCpU-tslsM",
        "outputId": "6c0846f4-6592-4747-95dc-b7ab2301ae5f"
      },
      "outputs": [],
      "source": [
        "for file in uploaded:\n",
        "  im = cv2.imread(file)\n",
        "\n",
        "  # Reset visualization threshold\n",
        "  output_score_threshold = 0.3\n",
        "  for cascade_stages in range(len(predictor.model.roi_heads.box_predictor)):\n",
        "    predictor.model.roi_heads.box_predictor[cascade_stages].test_score_thresh = output_score_threshold\n",
        "\n",
        "  # Run model and show results\n",
        "  outputs = predictor(im)\n",
        "  v = Visualizer(im[:, :, ::-1], metadata)\n",
        "  out = v.draw_instance_predictions(outputs[\"instances\"].to(\"cpu\"))\n",
        "  cv2_imshow(out.get_image()[:, :, ::-1])"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [],
      "name": "Detic_demo.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "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": "3.8.6"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}