How to make ground truth data
ljquality opened this issue · comments
We want to make a crack detection dataset and want to know how to make ground truth data (methods ， softwares or tools). Crack ground truth is different from ordinary segmentation ground truth. The shape of crack is so slim to draw contours, while ordinary object, such as human and cat, is easy to draw contours. Thankyou very much!
我们想了解一下如何去做一个裂缝的标注（方法、涉及的软件或者工具等） 。我们知道，裂缝形状很细长， 很难像普通物体一样使用labelme等工具去标注。所以，我们想请问一下你们有没有标注裂缝的比较好的方式。很抱歉，在这个数据集开放了几年后，问这个问题。十分感谢。
@ljquality 你好 我想请问对于你所提出的问题找到了解决方法吗
@ljquality @little1TouwaErio 请问你们知道怎么去做标签了吗
不知道呢，同求
import scipy.io
import numpy as np
import cv2
mat = scipy.io.loadmat('001.mat')
np_seg = mat['groundTruth'][0][0][0]
(y, x) = np.where(np_seg == 2)
np_seg[y, x] = 255
(y, x) = np.where(np_seg == 1)
np_seg[y, x] = 0
cv2.imwrite('./001.png', np_seg)
print()
We want to make a crack detection dataset and want to know how to make ground truth data (methods ， softwares or tools). Crack ground truth is different from ordinary segmentation ground truth. The shape of crack is so slim to draw contours, while ordinary object, such as human and cat, is easy to draw contours. Thankyou very much!
我们想了解一下如何去做一个裂缝的标注（方法、涉及的软件或者工具等） 。我们知道，裂缝形状很细长， 很难像普通物体一样使用labelme等工具去标注。所以，我们想请问一下你们有没有标注裂缝的比较好的方式。很抱歉，在这个数据集开放了几年后，问这个问题。十分感谢。
请问您知道对这类线状目标怎么标注了 么？
同求
We want to make a crack detection dataset and want to know how to make ground truth data (methods ， softwares or tools). Crack ground truth is different from ordinary segmentation ground truth. The shape of crack is so slim to draw contours, while ordinary object, such as human and cat, is easy to draw contours. Thankyou very much!
我们想了解一下如何去做一个裂缝的标注（方法、涉及的软件或者工具等） 。我们知道，裂缝形状很细长， 很难像普通物体一样使用labelme等工具去标注。所以，我们想请问一下你们有没有标注裂缝的比较好的方式。很抱歉，在这个数据集开放了几年后，问这个问题。十分感谢。请问您知道对这类线状目标怎么标注了 么？
同求
用普通分割方式标注多边形就可以，标细一点
We want to make a crack detection dataset and want to know how to make ground truth data (methods ， softwares or tools). Crack ground truth is different from ordinary segmentation ground truth. The shape of crack is so slim to draw contours, while ordinary object, such as human and cat, is easy to draw contours. Thankyou very much!
我们想了解一下如何去做一个裂缝的标注（方法、涉及的软件或者工具等） 。我们知道，裂缝形状很细长， 很难像普通物体一样使用labelme等工具去标注。所以，我们想请问一下你们有没有标注裂缝的比较好的方式。很抱歉，在这个数据集开放了几年后，问这个问题。十分感谢。请问您知道对这类线状目标怎么标注了 么？
同求用普通分割方式标注多边形就可以，标细一点
多谢回复，那进分割网络时候，图片是不是不能resize啊，尽量是多大的图就多大进网络？
还有就是我的数据集是电线，能用多边形标注么，多边形标注的话，是不是用lable me 就行？
We want to make a crack detection dataset and want to know how to make ground truth data (methods ， softwares or tools). Crack ground truth is different from ordinary segmentation ground truth. The shape of crack is so slim to draw contours, while ordinary object, such as human and cat, is easy to draw contours. Thankyou very much!
我们想了解一下如何去做一个裂缝的标注（方法、涉及的软件或者工具等） 。我们知道，裂缝形状很细长， 很难像普通物体一样使用labelme等工具去标注。所以，我们想请问一下你们有没有标注裂缝的比较好的方式。很抱歉，在这个数据集开放了几年后，问这个问题。十分感谢。请问您知道对这类线状目标怎么标注了 么？
同求用普通分割方式标注多边形就可以，标细一点
多谢回复，那进分割网络时候，图片是不是不能resize啊，尽量是多大的图就多大进网络？
还有就是我的数据集是电线，能用多边形标注么，多边形标注的话，是不是用lable me 就行？
我建议能用最新的就用最新的标注工具如https://github.com/ucbdrive/scalabel
We want to make a crack detection dataset and want to know how to make ground truth data (methods ， softwares or tools). Crack ground truth is different from ordinary segmentation ground truth. The shape of crack is so slim to draw contours, while ordinary object, such as human and cat, is easy to draw contours. Thankyou very much!
我们想了解一下如何去做一个裂缝的标注（方法、涉及的软件或者工具等） 。我们知道，裂缝形状很细长， 很难像普通物体一样使用labelme等工具去标注。所以，我们想请问一下你们有没有标注裂缝的比较好的方式。很抱歉，在这个数据集开放了几年后，问这个问题。十分感谢。请问您知道对这类线状目标怎么标注了 么？
同求用普通分割方式标注多边形就可以，标细一点
多谢回复，那进分割网络时候，图片是不是不能resize啊，尽量是多大的图就多大进网络？
还有就是我的数据集是电线，能用多边形标注么，多边形标注的话，是不是用lable me 就行？我建议能用最新的就用最新的标注工具如https://github.com/ucbdrive/scalabel
好的，多谢你 了
I also encountered a similar labeling problem, and I tried some tools that were not very efficient. I chose photoshop in the end, and the magic wand tool is actually quite easy to use.
I also encountered a similar labeling problem, and I tried some tools that were not very efficient. I chose photoshop in the end, and the magic wand tool is actually quite easy to use.
Obviously labelme is not very suitable for crack labeling. CVAT can label semi-automatically, but the target is too small, the pixel width in many places is too small, and the edge labeling is still not very clear and convenient.
I spent a lot of time on labeling with cvat, and finally found that using the magic wand tool of ps is more efficient. With a little processing, it can be made into labels for training. Share with you, if there is any reference value.