--- --- # obj on web # 參考資料 [Tensorflow Object Detection API Web Service](https://github.com/webrtcHacks/tfObjWebrtc) https://github.com/SamsungInternet/ar-demos # 基本安裝 ```shell= sudo apt-get update sudo apt-get install tasksel sudo tasksel cd /var/www/html ``` ### git clone 參考資料 ```shell= sudo git clone https://github.com/webrtcHacks/tfObjWebrtc.git cd tfObjWebrtc/ ``` # Install docker ```shell= sudo apt-get update sudo apt-get install \ apt-transport-https \ ca-certificates \ curl \ gnupg-agent \ software-properties-common curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add - sudo apt-key fingerprint 0EBFCD88 sudo add-apt-repository \ "deb [arch=amd64] https://download.docker.com/linux/ubuntu \ $(lsb_release -cs) \ stable" sudo apt-get update sudo apt-get install docker-ce docker-ce-cli containerd.io apt-cache madison docker-ce sudo docker run hello-world ``` # 進入root 創建tf-webrtchacks container ## 將container的5000端口映射到主機的5000端口 ```shell= sudo docker run -it -p 5000:5000 --name tf-webrtchacks -v $(pwd):/code chadhart/tensorflow-object-detection:webrtchacks ``` ## 解決運行server.py的問題 ```shell= python setup.py install protoc object_detection/protos/*.proto --python_out=. apt-get update apt-get install wget wget -O protobuf.zip https://github.com/google/protobuf/releases/download/v3.0.0/protoc-3.0.0-linux-x86_64.zip unzip protobuf.zip ./bin/protoc object_detection/protos/*.proto --python_out=. export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim ``` ## 把http變https(生產憑證) ```shell= wget https://ftp.openssl.org/source/openssl-1.0.2r.tar.gz //ssl生產工具 tar -zxvf openssl-1.0.2r.tar.gz cd openssl-1.0.2r/ apt-get install build-essential ./config --prefix=/usr/local --openssldir=/usr/local/openssl make make test make install openssl version //版本 which openssl //openssl的位置 openssl req -x509 -newkey rsa:4096 -nodes -out cert.pem -keyout key.pem -days 365 //生產憑證 nano server.py //最下面有個with SSL，改成下面程式碼 app.run(debug=True, host='0.0.0.0') 將此行註解 app.run(debug=True, host='0.0.0.0', ssl_context=('cert.pem','key.pem')) ``` # 換model * 下載[zip](https://drive.google.com/file/d/1RbkCuZSkIzByKfPzqxQrbFK-d16WjTae/view?usp=sharing)檔 * 上傳訓練後的 frozen_inference_graph.pb 、 labelmap.pbtxt 、 object_detection_api.py * 替換 ssd_mobilenet_v1_coco_2017_11_17 內的 frozen_inference_graph.pb * 將 labelmap.pbtxt 複製到 object_detection/data * 替換 object_detection_api.py 檔案 ```shell= exit sudo rm object_detection_api.py cd ssd_mobilenet_v1_coco_2017_11_17 sudo rm frozen_inference_graph.pb cd sudo cp frozen_inference_graph.pb /var/www/html/tfObjWebrtc/ssd_mobilenet_v1_coco_2017_11_17 sudo cp labelmap.pbtxt /var/www/html/tfObjWebrtc/object_detection/data sudo cp object_detection_api.py /var/www/html/tfObjWebrtc/ cd /var/www/html/tfObjWebrtc/ ``` ## 會遇到的狀況 * No op named GatherV2 in defined operations. * Tensorflow的版本不合『docker的Tensorflow的版本 != 訓練時的Tensorflow的版本』 ## 要改docker下面的Tensorflow的版本 * 改成你訓練時的Tensorflow的版本，pip install tensorflow==XXX ```shell= sudo docker rm tf-webrtchacks sudo docker run -it -p 5000:5000 --name tf-webrtchacks -v $(pwd):/code chadhart/tensorflow-object-detection:webrtchacks pip install tensorflow==1.9.0 python server.py ``` ## 執行'object_detection_api_test.py'會遇到這個問題 * Attempted to use a closed Session * 參考 https://www.twblogs.net/a/5b8c36e52b71771883315b6d * session 執行後會關閉,導致server.py內的程式碼get_objects的function 沒辦法執行所以要把session的程式碼改到get_objects內 * 改object_detection_api.py 這個檔案內的get_objects # 套用 3D 圖檔 ```shell= cd /var/www/html/tfObjWebrtc/static sudo git clone https://github.com/SamsungInternet/ar-demos.git sudo git clone https://github.com/jphone0921/3d.git sudo rm objDetect.js sudo rm local.html cd /var/www/html/tfObjWebrtc/static/3d sudo cp objDetect.js .. sudo cp local.html .. ``` # 後製鏡頭 參考資料 https://developer.mozilla.org/en-US/docs/Web/API/MediaDevices/getUserMedia * local.js 的video改成以下 ```shell= video: { width: {min: 640, ideal: 1280, max: 1920}, height: {min: 480, ideal: 720, max: 1080}, facingMode:"environment" } ``` ## 改鏡頭左右相反問題 * 修改 local.html 中的 -webkit-transform: scale(1, 1); 參考資料 https://tympanus.net/codrops/css_reference/scale/ # 運行 * 要先移除docker * 在執行docker ```shell= python server.py http://localhost:5000/local chrome需https (火狐不用) http://localhost:5000/video 狗狗 ``` ```shell= sudo docker rm tf-webrtchacks cd /var/www/html/tfObjWebrtc/ sudo docker run -it -p 5000:5000 --name tf-webrtchacks -v $(pwd):/code chadhart/tensorflow-object-detection:webrtchacks pip install tensorflow==1.9.0 python server.py ``` # 結果 當辨識到撲克牌的時候，會相對應的出現其數字而在螢幕上顯示POKEMON。 EX: 數字9 > 皮卡丘