Closed slimhappy closed 3 years ago
Hello, tancik ! I am keen on your outstanding work. But when I was traing a detector for distiguishing the encoded images from complex background (not totally white), I have some trouble about making dataset. Here is my questions:
- Are the encoded images placing on totally white background in the traing dataset for detector?
- If the encoded image's size is small (maybe 1/100 of an A4 paper), how can I make a training dataset to get great precision ? Finally, I have written a documentation to guide chinese novices to build an environment and turn your code from tensorflow 1.x to tensorflow 2.x. I hope you will like it and it can help more githubers to run your work.
Hey, is this problem solved?@slimhappy
Hello, tancik ! I am keen on your outstanding work. But when I was traing a detector for distiguishing the encoded images from complex background (not totally white), I have some trouble about making dataset. Here is my questions:
- Are the encoded images placing on totally white background in the traing dataset for detector?
- If the encoded image's size is small (maybe 1/100 of an A4 paper), how can I make a training dataset to get great precision ? Finally, I have written a documentation to guide chinese novices to build an environment and turn your code from tensorflow 1.x to tensorflow 2.x. I hope you will like it and it can help more githubers to run your work.
Hey, is this problem solved?@slimhappy
Not yet
Hello, I received a reply from the author as follows:
'' I don't remember exactly how large the transformations were, I just chose a value that matched what I wanted to recover in the real-world. In fact the detector is just learning how to pick out an image on another image. I found that training it on just square images (not stegastamped) pasted on a large image worked just as well. Because of the error detection when decoding the stegastamp it is fine if the detector isn't perfect. There is a lot that could be explored to make the detector better including end-to-end training. I am excited to see if someone pursues this direction. ''
You can have a @.***
------------------ 原始邮件 ------------------ 发件人: "tancik/StegaStamp" @.>; 发送时间: 2021年4月23日(星期五) 中午11:17 @.>; @.**@.>; 主题: Re: [tancik/StegaStamp] Some questions about making detector dataset (#26)
Hello, tancik ! I am keen on your outstanding work. But when I was traing a detector for distiguishing the encoded images from complex background (not totally white), I have some trouble about making dataset. Here is my questions:
Are the encoded images placing on totally white background in the traing dataset for detector?
If the encoded image's size is small (maybe 1/100 of an A4 paper), how can I make a training dataset to get great precision ? Finally, I have written a documentation to guide chinese novices to build an environment and turn your code from tensorflow 1.x to tensorflow 2.x. I hope you will like it and it can help more githubers to run your work.
Hey, is this problem @.***
Not yet
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.
Hello, tancik ! I am keen on your outstanding work. But when I was traing a detector for distiguishing the encoded images from complex background (not totally white), I have some trouble about making dataset. Here is my questions:
If the encoded image's size is small (maybe 1/100 of an A4 paper), how can I make a training dataset to get great precision ?
Finally, I have written a documentation to guide chinese novices to build an environment and turn your code from tensorflow 1.x to tensorflow 2.x. I hope you will like it and it can help more githubers to run your work.