fitzpchao / bandon Goto Github PK

你好前辈，我看您的BT-flow标签里面是.tif格式的图像，32位深度的图，我想问一下，怎么解析这个图像里面的信息呢，因为我的数据集也想制作一下这种偏移的标签数据，可以给我一些建议吗，谢谢

Thank you for your outstanding work. Your dataset includes images from multiple perspectives and dates within the same area. May I ask if the images in this dataset have undergone orthorectified correction? In other words, does DEM perform orthorectified correction on each image?

您好，我使用ps软件打开了.tif格式的标签，但是里面显示的部分有些地方有一些歧义，如L81_00639_783959这张图，打开之后有三个不同的通道，里面显示了不同的内容
下面分别是三个不同通道的图像，这个是什么问题呢

您的数据集一共有四种真值，ST-offsets，Seg，BI-flows，Final change mask这四种，我想问一下，您如果想对比建筑物之间是否发生变化，是不是只需要Seg，BI-flows，Final change mask这三种就可以了，是不是就不需要ST-offsets也可以了，因为我们找到对应建筑物的目的已经到了不是吗？期待您的回复谢谢

您好，我看的您的代码里面MTGCDNet.py里面，num_classes分别是2，3，3，10，10，10个类别。2指的是最终的变化图，3指的是分割结果图，10是ST-flow和BI-flow图，是这样吗？那么您是用的分类的方法去做的偏移嘛。我这样理解对吗

After delving into your recent research papers, I was deeply impressed by the remarkable contributions you have made to the field. The moment your source code became open-source, I eagerly embarked on a journey to learn from it and attempt to reproduce the exceptional model performance described in the papers. However, during the process of debugging and fine-tuning the code, I encountered significant discrepancies in the model's performance compared to the reported results. I am reaching out with a sincere and humble request for guidance and assistance. I think there might be subtle intricacies or specific procedures that I might have overlooked while trying to reproduce your work. Therefore, I was hoping you could give me some advice, tips or clarification about the code implementation or any potential issues I should be aware of, thank you very much!

Hi, I cannot run the code with the python libraries version in your instructions. Would you like to provide the environment requirements for this repo?