我想使用Faster-RCNN方法在Tensorflow中进行对象检测。
我计划使用Amazon Mechanical Turk众包来获取注释。但是,有没有人知道是否有任何支持mturk的边界框注释工具并以Pascal VOC格式在xml文件中输出注释。
答案 0 :(得分:0)
这里是详细的文档和源代码: https://github.com/trinath503/Python/tree/master/Generate_Pascal_VOC_Files
注意:您需要以指定的格式传递数据
pascal_voc_data = '''
[
{
"folder":"folder",
"filename": "1.jpg",
"path":"path",
"source":{"database":"database"},
"size":{"width":256,"height":256,"depth":3},
"segmented":0,
"objects":[{"name":"name","pose":"pose","truncated":"truncated","occluded":"occluded","bndbox":{"xmin":3,"xmax":33,"ymin":3,"ymax":33}},{"name":"name","pose":"pose","truncated":"truncated","occluded":"occluded","bndbox":{"xmin":3,"xmax":33,"ymin":3,"ymax":33}}]
},
{
"folder":"folder",
"filename": "2.jpg",
"path":"path",
"source":{"database":"database"},
"size":{"width":256,"height":256,"depth":3},
"segmented":0,
"objects":[{"name":"name","pose":"pose","truncated":"truncated","occluded":"occluded","bndbox":{"xmin":3,"xmax":33,"ymin":3,"ymax":33}},{"name":"name","pose":"pose","truncated":"truncated","occluded":"occluded","bndbox":{"xmin":3,"xmax":33,"ymin":3,"ymax":33}}]
}
]
'''