Features

GSense provides 3 main functionalities for the analysis and annotation of hyperspectral images - spectral indexing, image segmentation, and label binarization.

Spectral Indexing

Lets you compute pseudo-RGB images from spectral bands by referencing them in mathematical expressions using channel indices.

Example of acceptable expressions:

Red: ch[10] + ch[20]/4 + 3/ch[2]
Green: ch[50]
Blue: ch[2] - ch[6]

Note: If an image in the loaded batch does not have the referenced channel data, pseudo-RGB image computation for that image is skipped.

Image Segmentation

Segment generated pseudo-RGB images using Meta AI's Segment Anything Model. Choose your model type and default or custom fine-tuned checkpoint to run segmentation with.

Binarizer

We use SAM to generates all masks and the label binarizer helps reduce the labels to binary instead.