Herschel vision: A hyperspectral image processing software for data preparation in machine learning pipelines

Abstract

In recent years, publications have relied on either commercial software or custom-developed codes for the analysis of hyperspectral images. This practice has inadvertently restricted the broader adoption of hyperspectral imaging. Herschel Vision is an open-source hyperspectral image analysis application that enables the users to visualize and analyze hyperspectral images. Herschel Vision can automate white and dark reference calibration, spectral data visualization and extraction, generation of pseudo RGB images, cropping and segmentation of region of interests, saving images as three-dimensional cubes, and applying forward feed signal pre-processing corrections. Herschel Vision can handle proximally recorded hyperspectral data for curation of machine learning and deep learning training datasets.