|
|
|
||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
CANNIBAL
Unveils the Hidden Gems: Hyperspectral Band Selection via Clustering of
Weighted Variable Interaction Graphs |
|
|||||
|
|
|
|
|
|
|
||
|
|
Abstract |
|
|
|
|
||
|
|
This seminar introduces CANNIBAL, a band selection algorithm designed
to address challenges posed by the high dimensionality of hyperspectral
images. Leveraging unsupervised clustering of inter-band dependencies
captured in weighted Variable Interaction Graphs, CANNIBAL optimizes band
selection for downstream tasks such as hyperspectral unmixing
and segmentation. Experimental results demonstrate its superiority over
existing algorithms, enabling significant reduction in the number of bands
without compromising model quality. Notably, CANNIBAL offers flexibility,
catering to both parametric and non-parametric use cases, thereby presenting
a promising solution for efficient analysis of hyperspectral data in various
fields. Time&Place 2024.04.11
09:15, Google meet Zoom: https://meet.google.com/wyd-akjx-xfr |
|
|||||
|
|
Tutorial
length |
1.5 hours |
|
|
|
|
|
|
|
Tutorial
level |
introductory |
|
|
|
|
|
|
|
Download |
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
