Qing
Some post-process that should only be applied to the SD model affect the LaMa result
Thank you for sharing the experiment detail! I am currently trying to implement the document dewarping method to achieve the following effect (recorded from https://www.textin.com/experience/text_auto_removal) https://user-images.githubusercontent.com/3998421/185350667-931bbd43-57b4-456f-9361-d2b627a051e3.mov Among several methods, PaperEdge...
As far as I know, maybe the closest approach to self-supervision is [Fourier Document Restoration for Robust Document Dewarping and Recognition](https://arxiv.org/abs/2203.09910). Although it open-sources the dataset, unfortunately, the authors do...
Doc3D 的数据太大了。。。还是等作者服务器恢复吧
> > Doc3D 的数据太大了。。。还是等作者服务器恢复吧 > > 您好，请问一下，数据量这么大，你们是怎样下载下来的，电脑的存储容量不够啊，还有就是能不能用其他数据集代替呢 原来就是通过作者的服务器下载的，据我所知没有这么全的数据集了，另一个选择是使用作者的代码自己生成 https://github.com/sagniklp/doc3D-renderer
With the new version (0.33.0) stable diffusion inpainting need 2GB for 512x512 image. Please check https://lama-cleaner-docs.vercel.app/memory#paint-by-example
Open the ngrok public URL(in 2. Setup ngrok) to access the web app
- Using `lama` model would be helped(MAT model will pad input to square, which require more VRAM) - Reduce the mask area( From the screenshot it looks like your mask...
With the new version (0.33.0) stable diffusion inpainting need 2GB for 512x512 image. Please check https://lama-cleaner-docs.vercel.app/memory to see other memory saving method
Is the model already downloaded? I suggest trying to start the server with the model you want to switch, to make sure it finishes the download