step1:開始

mkdir Github

cd Github

git clone https://github.com/bruce601080102/yolov7_stream_rtc.git

cd yolov7_stream_rtc/docker

./build.sh

step2:進入容器內

docker exec -it mydeepstream:v2 bash

cd /opt/nvidia/deepstream/deepstream-6.1/commonFolder/Github/yolov7_stream_rtc

pip3 install -r requirement.txt

pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113

step2-1:內置zsh

sh -c "$(curl -fsSL https://raw.githubusercontent.com/bruce601080102/yolov7_stream_rtc/main/docker/build_zsh.sh)" 

step3:pt下載

cd /opt/nvidia/deepstream/deepstream-6.1/commonFolder/Github/yolov7_stream_rtc

mkdir stream/yolo/weight/default

cd stream/yolo/weight/default

wget https://github.com/WongKinYiu/yolov7/releases/download/v0.1/yolov7.pt

串流執行

python3 -m streamlit run navigation.py --server.port 9998 --server.enableCORS=false

本地網址:http://localhost:9998

ssl憑證測試

• 由於攝像頭啟動基於安全需要憑證或是localhost域名才能順利執行,因此可以使用proxyssl來測試是否能正常使用

cd ssl
./ssl-proxy-linux-amd64 -from 192.168.10.112:9998 -to 192.168.10.112:8502

ssh綁網域設定測試

export STREAMLIT_RUN_FILE_OR_URL=navigation.py

step4:Tensor-RT生成

• 執行以下命令集會生成出engine文件,一個環境只能生出對應的engine,因此更換環境需要重新生成
• zsh環境下無法使用,需切回bash

cd stream/output_rt

chmod u+x *.sh

mkdir engine

./build.sh

step5:supervisorctl啟動

supervisorctl start all

DEMO

Default Demo	PlateLicense Demo