labelme vs others
Labelme
labelme: qt based passed
Labelme-web
web-based(参考资料简单、少)
github code: https://github.com/CSAILVision/LabelMeToolbox
https://github.com/CSAILVision/LabelMeAnnotationTool
usage:https://blog.roboflow.com/labelme/
step1: 注册账号并登陆 http://labelme2.csail.mit.edu/Release3.0/browserTools/php/browse_collections.php?username=hayley&folder=
step2: 上传图片到labelme(缺点:一次最多上传20张,只支持jpg,图片大小有限制5M以下)
step3: 开始标注(生成的标注信息是xml格式)
step4: 从labelme下载数据(将图片和标注压缩成zip包)
step5: 将上一步下载的解压,上传到lroboflow(需要注册)中。可以将标注转换成别的模型识别的格式。或者对数据进行预处理等,也可以用里面的模型对数据进行。 页面风格和tusai的风格不是很搭
semantic-segmentation-editor
web-based meteor app, react,paper.js, three.js 最新版本1.5.3 提供了docker image
使用步骤:
step1: 安装并启动.默认端口为80
wget https://raw.githubusercontent.com/Hitachi-Automotive-And-Industry-Lab/semantic-segmentation-editor/master/sse-docker-stack.yml
SSE_IMAGES=/Users/hayley/Desktop/semantic-segmentation-editor/images
docker-compose -f sse-docker-stack.yml up
met error
ERROR: for semantic-segmentation-editor_app_1 Cannot create container for service app: create .: volume name is too short, names should be at least two alphanumeric characters
fix:创建data目录,并且修改volumes值
volumes:
- "${PWD}/images/:/root/sse-images:rw"
step2: 打开浏览器http://localhost/browse/0/20/ step3: 标注,生成json标注信息
VIA
- http://www.robots.ox.ac.uk/~vgg/software/via/via.html
- http://www.robots.ox.ac.uk/~vgg/software/via/