-
Notifications
You must be signed in to change notification settings - Fork 19.4k
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
Conv1D on custom shaped data #19900
Comments
Hi @Wasim04 - Healpix stands for Hierarchical Equal Area isoLatitude Pixelization of a sphere where partitioning sphere at higher resolutions. So this partition of sphere surface has different size pixels and which is represented as 2D array. So it is better to use Conv1D and Conv2D both are not suitable for healpix image data. Conv1D can create convolution on single dimension data while Conv2D can create convolution on two dimensional data. JAX library is better option to implement spherical CNNs. Definitely we can do experiment with Conv1D with healpix data(by projecting MNIST dataset onto HEALPix), but it will not generate good result for large spherical data. Let me know if you required more details. Thanks..!! |
This issue is stale because it has been open for 14 days with no activity. It will be closed if no further activity occurs. Thank you. |
@mehtamansi29 Thanks for your suggestion. Healpix by design is a 1D array of pixels. Each pixel can carry some value to provide a global localised representation of some data (atmospheric or astrophysical etc) since we can use Theta and Phi as pixel coordinates. The problem arises when these pixels are oriented in a particular way. That is, nested or ring, that I was wondering in my first query. Because a Conv1D will construct a kernel which will be 1D array likey in squared pixel shape and the stride will guide the kernel movement. I am looking for some suggestions if there is a way to address this pixel orientation in the kernel of a traditional Keras Conv1D. Essentially a kernel that follows a nested or ring scheme. Many Thanks |
This issue is stale because it has been open for 14 days with no activity. It will be closed if no further activity occurs. Thank you. |
Hi,
Just wondering if there is a way to use Conv1D in a heapix data. That is, it can handle, resolution(nside) and pixel orientation (nested or ring).
Many Thanks
Wasim
The text was updated successfully, but these errors were encountered: