LeaRun
LeaRun
When the code runs to image = [x["iamge"].to(self.device) for x in batch_inputs] Code error “too many indices for tensor of dimension 3”
> No I need a pretrained model. I don't have data - Sent from my phone. Pls excuse typos & brevity. > […](#) > On Sun, Oct 22, 2023, 6:41...
> Which one? What is the effect of using this model to train images in non remote sensing fields 
> @wangqingl 嗨，我想我有一些在其他领域使用更改检测算法/模型的经验。我认为最困难的是变化检测数据集中的图像对齐得很好。因此，在变化检测问题中，您不需要在图像预处理中意识到对齐问题。如果你的图像对齐得很好，我认为值得对你的数据集进行微调。 Okay, thank you for your reply. What open-source methods have you used in non remote sensing datasets? Can you share with me? The alignment error of my...
> @wangqingl I using ChangeFormer: https://github.com/wgcban/ChangeFormer/tree/main. It is also including in this repo. What open-source methods you already applied? Thank you, but I only use this method currently.
> 训练是否正常呢？  你好，训练的时候报了这个错误，我查了一下，有人说安装yapf==0.40.1版本，但是我换了这个版本还是有这个错误
> 请截个完整的图 
不好意思，是我搞错了，现在跑起来了，那个方法可以解决问题
标签是单通道吗，但是我下载了cdd的数据集的标签，也是三通道的  > @wangqingl 这不是写的很明显吗，标签格式不对，应该是单通道的
> @wangqingl 这不是写的很明显吗，标签格式不对，应该是单通道的 好的我知道了，谢谢你