chail / anyres-gan Goto Github PK
View Code? Open in Web Editor NEWLicense: Other
License: Other
I notice the example in #quickstart-with-pretrained-models can only generates square images.
I modify the code, let width and height be different, it will not work.
Of course, this question is not about panorama training.
Is it limited by the architecture of stylegan3 ?
support an easier way to prepare dataset,please
Is there an easy way to do both scaling and 360-degree panorama at the same time? It would be great to generate high resolution panoramas.
Great work!
Hi,
First of all thank you for the amazing work. I've been trying the interactive demo and it is really impressive. I'd like to use it for face super-resolution, for a project of the company I'm currently working for.
Now I'd like to enhance resolution of 256x256 images I have but I am not sure it is possible. Besides, in the paper, you say you are not a super resolution model. And indeed, in the interactive colab demo, we have to sample z and map it to ws, which is impossible if we want to use an already existing image, right ?
Is it the same issue/question as conditional generation ? In another issue about this topic you said that the framework could be modify to do so. Could you please give additional information about the necessary steps and changes to make in order to do so and use your model as a super resolution module ?
Edit : after checking again, conditional generation is more like condition to a class, an attribute of something like this, right ? My question is more about how to use any-resGAN as a super resolution module to enhance resolution of frames of a video ?
Thanks a lot !
Geoffrey
Hey first thank you so much for releasing this work with all of the code, very impressive! I'm trying to download the pretrained models using the download_resources.sh
script. However, I'm running into an issue where the model file seems to be missing or removed from the server, am I trying the right URL?
Thanks again!
I am trying to train this model on a new dataset. I noticed that the prompt in stage2 said "the teacher can be any sgan3 pretrained model, e.g. from the stylegan3 official repository also works", so I downloaded the stylegan3 pre-training model on NVIDIA catalog(https://catalog.ngc.nvidia.com/orgs/nvidia/teams/research/models/stylegan3), but when it was used as the stage2 teacher, an error was reported that the size of the Tensor did not match, I want to know why this error occurred? Am I got it wrong?
Sorry to take up your time. I'm looking forward to your reply.
Hi,
Interesting work! Can you an approximate date by which you will release the code?
Thanks,
Sonam
thanks in advance.
I am working on generating conditional food images but met some errors on loading labels for patch data. I strictly follow the instructions in the file "train.sh" and "train.py". In the first stage of training, I first tried to leverage the pre-trained network of conditional stylegan3 to train the patch datasets in stage two and I got the following errors:
And it is caused by None type tensor as you can see from the terminal output messages.
I read the paper, the authors use the unconditional image generation method. But I want to know if this method is compatible with conditional image generation.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.