linhandev / plane

plane

IFS=$'\n'

for folder in `ls`
do
echo $folder
cd $folder
ffmpeg -framerate 8 -pattern_type glob -i "*.png" ${folder}.mkv
mv ${folder}.mkv ../../mkv
cd ..
# rm $folder -rf
done

dir="Q23502-撤轮挡"
cd $dir
ffmpeg -framerate 2 -pattern_type glob -i "*.png" ${dir}.mkv
mv $dir.mkv ../../mkv
cd ..
rm -rf $dir

mkdir ../mp4
for f in `ls`
do
  cd $f
  mv * ../../mp4
  cd ..
done

for f in `ls`
do
  f=${f%????}
  echo $f
  cp /home/lin/Desktop/data/plane/flg/ann/train/flg/${f}.xml /home/lin/Desktop/data/plane/flg/ann/train/
done

python det/f_flg.py -i /home/lin/Desktop/data/plane/frame/img-temp/ -o /home/lin/Desktop/data/plane/frame/xml-temp/ --bs 8 --model model/best_model/

paddlehub 这个yolo的可视化有bug，一个batch的数据都会用第一张图片之后往上画框 TODO：给这个问题pr

注意： 所有数字要补0,否则排序很费劲 一个文件名不要用两个同样的分隔符

sudo docker run -d -it -p 6666:8080 -v /home/hanlin/label-studio/docker-data/:/label-studio/data heartexlabs/label-studio:latest

了docker exec -it c460 /bin/bash echo "TASKS_MAX_FILE_SIZE = 9999999999999999" >> /label_studio/core/settings/label_studio.py docker logs -f

docker system prune -a # 删除所有本地存的image

systemctl --user start ml journalctl --user -f -u ml