-
Notifications
You must be signed in to change notification settings - Fork 30
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
data clues and research usage #4
Comments
Hi! Thanks for your interest! The training data consists of a few dozens of images with all the stars deleted in PS by hand (there is no other way ensuing high quality of star removal and leaving all other areas of images exactly as they are). If you want to use it for your particular data, you can clean a few images of your own data and fine-tune existing weights so it works better for your particular star profiles. Unfortunately, I am not publishing my own training data under any conditions since that's where the huge amount of work went, but I am available to help you as much as I can. Let me know if you are interested and we can discuss things in more details! |
Thanks @nekitmm for your help! We discuss it within our team (and probably try to prepare our own images) and I will return with more questions if needed. |
Thanks for reaching out, @Amarchuk! Please do not hesitate to reach out to me with any questions! I am actually very interested in how this will turn out and available for active participation in your project if you want! I am myself in academia and if this will turn into some scientific publication, that will be awesome! |
Awesome! What are your research interests? It seems like for now creation of Photoshop-corrected images is the hardest part of preparation (there can be literally hundreds of stars on the image and I don't know yet if it is possible to remove them). How you manage to do that? Any tricks or advices are highly appreciated. |
@Amarchuk, lets take this conversation elsewhere |
Hello,
I am an astrophysicist from team working with University of Waterloo (UW), Canada. We find this project very interesting and of potential usage for our big ongoing full-sky survey. We would like to experiment with improving
starnet
performance a little (because unfortunately it is not always produce adoptable results) and probably tune it for FITS we will use. Could you share your expertise about data you train on? Specifically:In any case we will surely acknowledge and referencing your work
The text was updated successfully, but these errors were encountered: