search

danfenghong / IEEE_TNNLS_EGU-Net

Danfeng Hong, Lianru Gao, Jing Yao, Naoto Yokoya, Jocelyn Chanussot, Uta Heiden, Bing Zhang. Endmember-Guided Unmixing Network (EGU-Net): A General Deep Learning Framework for Self-Supervised Hyperspectral Unmixing, IEEE TNNLS, 2021.
56 stars 13 forks source link

Larger areas #3

Open RichardScottOZ opened 3 years ago

RichardScottOZ commented 3 years ago

Nicely done Danfeng, thanks for making this available.

Here's a question - you are using VCA as a preliminary (and presumably could use any sort of endmember extraction).

If wanting to do this over a large area this would be problematic - need to be parallelised - e.g. if you had billions of pixels (or more).

Do you think taking endmembers taken from a simpler neural network/ANN that could handle work at such a scale via patching would be a reasonable substitute?

danfenghong commented 3 years ago

Nicely done Danfeng, thanks for making this available.

Here's a question - you are using VCA as a preliminary (and presumably could use any sort of endmember extraction).

If wanting to do this over a large area this would be problematic - need to be parallelised - e.g. if you had billions of pixels (or more).

Do you think taking endmembers taken from a simpler neural network/ANN that could handle work at such a scale via patching would be a reasonable substitute?

Hi Richard, Thank you very much for your interest in our work! Your idea for larger areas is acceptable. You can separately address the spectral unmixing in parallel when facing a large area, e.g., including billions of pixes and more.

Thanks again!