You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
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..!!
@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.
Thanks for the jax suggestion, I shall look into it.
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: