--- title: 雲端技術實務期末專題 tags: 期末專題 --- # 手寫辯識 Docker 部署 :::spoiler 目錄 [TOC] ::: ## 環境 作業系統：Arch Linux KDE 文字編輯器：Sublime Text 4 虛擬化部署：Docker 24.0.2 虛擬機器平台：OpenStack 手寫辯識： - Tensorflow - OpenCV - Pillow 前端：p5.js 網頁框架：Flask ## 製作 Model * [Colab](https://colab.research.google.com/drive/1uqs9DOVtDc3nOFwBiJ6K51fknN5MZvwc?usp=sharing) ## 安裝 Docker * 更新 `keyring` ```bash $ sudo pacman -Syy archlinux-keyring ``` * 下載 `docker` ```bash $ sudo pacman -Sy docker ``` * 安裝套件三連 ```bash $ sudo systemctl enable docker ``` ```bash $ sudo systemctl start docker ``` ```bash $ sudo systemctl status docker ``` ![](https://hackmd.io/_uploads/H1wJwb98h.png) > Ative 顯示 actvie ( running ) 就代表安裝成功 * 開始下指令會噴錯 ![](https://hackmd.io/_uploads/ry6FwZcI3.png) * 設定群組 ```bash $ sudo groupadd -f dcoker $ sudo usermod -aG docker $USER $ newgrp docker ``` * 最後下這個會看到 `docker` 會在最前面 ```bash $ groups ``` ![](https://hackmd.io/_uploads/SyESK-58h.png) * `docker` 指令就可以正常下了 ![](https://hackmd.io/_uploads/B1NOFbcU2.png) * 測試 `docker` ```bash $ docker run hello-world ``` ![](https://hackmd.io/_uploads/ryJmqWqL2.png) > 要先 `docker pull hello-world` * 下載 `ubuntu 22.04`（[連結](https://hub.docker.com/_/ubuntu)） ```bash $ docker pull ubuntu:22.04 ``` ![](https://hackmd.io/_uploads/HJrksW5In.png) > 看到這樣代表下載成功 ## 製作 Image ### Dockerfile ```dockerfile FROM ubuntu:22.04 RUN apt-get update && apt-get install -y \ python3 \ python3-pip \ python3-dev \ && apt-get clean RUN pip3 install tensorflow RUN pip3 install opencv-python-headless RUN pip3 install flask RUN pip3 install pillow COPY model.h5 /app/cloud_model.h5 COPY static/main.js /app/static/main.js COPY static/style.css /app/static/style.css COPY templates/index.html /app/templates/index.html COPY app.py /app/app.py CMD python3 /app/app.py ``` ### 其他 Code 放在 [qaz5823091/NumberRecognition - Github](https://github.com/qaz5823091/NumberRecognition) ### 檔案結構 ![](https://hackmd.io/_uploads/HkfnD6ewh.png) ### 試跑 * build ```bash $ docker build -t number_recognition . ``` * run ```bash $ docker run nubmer_recogntion ``` ### 關閉 container * 查看 ID ```bash $ docker ps ``` * 再透過 `id` 停止 ```bash $ docker containter stop <id> ``` ### 打包 * 把 `image` 附上 `tag` ```bash $ docker tag number_recognition cpp1092020/web_dl_1092926 ``` * 再來 `push` ```bash $ docker push cpp1092020/web_dl_1092926 ``` * 結果 ![](https://hackmd.io/_uploads/H1yLB8kP3.png) ### 執行 * 把 `image` 載下來 ```bash $ docker pull cpp1092020/web_dl_1092926 ``` > 應該會跑很久，因為有 `tensorflow` * 最後執行 ```bash $ docker run -v ./static:/app/static -p 5000:5000 number_recogition ``` > 要先到路徑內 > `-v` 是把使用者丟的圖片對應到 Host Computer，應該也可以不用 ## 成果 * 輸入圖片 ![](https://hackmd.io/_uploads/SJxzVpxP3.png) * 手寫數字 ![](https://hackmd.io/_uploads/B1r44pew3.png) ## 參考 * [Day4：用簡單的例子來說明如何使用 Docker 指令](https://ithelp.ithome.com.tw/articles/10190921) * [Docker permission denied](https://phoenixnap.com/kb/docker-permission-denied) * [TensorFlow 2 教學：MNIST--自訂 Model 進階教學](https://weikaiwei.com/tf/tensorflow-2-mnist-2/) * [上傳 Docker Image 到 Docker Hub](https://ithelp.ithome.com.tw/articles/10192824) * [強化訓練](https://github.com/SweetornotspicyMarathon/Learn-Tensorflow-Keras-Note/blob/main/U4%2BU6%2BU7.ipynb) * [CV2 強化圖片](https://github.com/eesgetdegrees/format-your-own-images-to-match-MNIST-dataset/blob/main/image_format.py)