Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

训练过程中生成的Image #10

Open
happywt opened this issue Aug 11, 2024 · 12 comments
Open

训练过程中生成的Image #10

happywt opened this issue Aug 11, 2024 · 12 comments

Comments

@happywt
Copy link

happywt commented Aug 11, 2024

作者您好,我在使用CVC数据集训练SDSeg时,在logs下的文件夹中,在images里生成了许多图片。如图所示。请问,这些都是什么图片吗?您能解释一下吗?谢谢
f586eb7e5592509be4e19f206f2c64e3

@happywt
Copy link
Author

happywt commented Aug 11, 2024

为什么会有关于房子的图像呢?
b824194cb162bc9d54043f397e461ba0

@lin-tianyu
Copy link
Owner

作者您好,我在使用CVC数据集训练SDSeg时,在logs下的文件夹中,在images里生成了许多图片。如图所示。请问,这些都是什么图片吗?您能解释一下吗?谢谢 f586eb7e5592509be4e19f206f2c64e3

这些图片是训练过程中记录下来的log,他们会显示在tensorboard中,方便我们在训练过程中监视训练过程。(tensorboard会读取这些图片并显示)

@lin-tianyu
Copy link
Owner

为什么会有关于房子的图像呢? b824194cb162bc9d54043f397e461ba0

这些房子(教堂)图片的出现是因为SDSeg中的denoising部分使用了SD提供的预训练模型,而这个预训练的数据集正是一个教堂图片的数据集。

模型中使用这个预训练模型只是为了提供一个较好的初始化,对训练的结果不会有太大影响~

@happywt
Copy link
Author

happywt commented Aug 13, 2024

作者您好,我在使用CVC数据集训练SDSeg时,在logs下的文件夹中,在images里生成了许多图片。如图所示。请问,这些都是什么图片吗?您能解释一下吗?谢谢 f586eb7e5592509be4e19f206f2c64e3

这些图片是训练过程中记录下来的log,他们会显示在tensorboard中,方便我们在训练过程中监视训练过程。(tensorboard会读取这些图片并显示)

好的,谢谢你~

@happywt
Copy link
Author

happywt commented Aug 13, 2024

为什么会有关于房子的图像呢? b824194cb162bc9d54043f397e461ba0

这些房子(教堂)图片的出现是因为SDSeg中的denoising部分使用了SD提供的预训练模型,而这个预训练的数据集正是一个教堂图片的数据集。

模型中使用这个预训练模型只是为了提供一个较好的初始化,对训练的结果不会有太大影响~

好的,我明白了,谢谢你~

@QiyuanY
Copy link

QiyuanY commented Nov 18, 2024

请问progressive代表的是哪个过程

@lin-tianyu
Copy link
Owner

什么progressive?

@QiyuanY
Copy link

QiyuanY commented Nov 18, 2024

image
这里展示的结果中带有progressive,这个过程是指通过采样生成图像的过程吗

@lin-tianyu
Copy link
Owner

是的。强烈建议使用tensorboard来观察训练过程,会非常清晰。

不过请注意,这里的progressive是不使用任何采样器,逆采样1000步的过程中的图像,而并不是本文所使用的单步采样方式。在log中记录这个只是为了方便对比和debug。

@hello123843
Copy link

image
image
Could you please explain in detail the meaning of these images? Specifically, I would like to confirm which ones are the images generated during the training process, starting from random noise sampling and gradually denoising. I would greatly appreciate your response.

@lin-tianyu
Copy link
Owner

Hi,

Thanks for your question. These images are not designed to be checked like this. They are stored here for tensorboard visualization. So you should start a tensorboard session to see those images.

Here is an example of how you can start a tensorboard session:

tensorboard --logdir= ./logs/

Here are some example of using tensorboard for monitoring the training process:
image
image

Have fun!

@lin-tianyu
Copy link
Owner

And btw, you'll find images of prediction starting from random noise sampling and gradually denoising in tensorboard, but those are just for visualization and comparison.

The highlight in SDSeg is that it only need one step of reverse process to generate the segmentation result, thus there is no gradual denoising process for that.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants