HSIToolbox
A software suite and a web application for labelling and managing HSI datasets, and training different deep learning models on these datasets.
Description:
This software aims to bridge the shift from model-centric to the data-centric deep learning approach by allowing both deep learning experts and domain experts to create impact in the real world. It provides a web-based user-friendly graphical interface that allows a domain expert to add, label and manage HSI (HyperSpectral Images) datasets, and train different out-of-the-box deep learning models on them. The software can read several dataset formats, so that hyperspectral images from different datasets can be uploaded in a uniform fashion for further labelling and classification. In addition, the software can run on a server with dedicated hardware for deep learning (i.e. GPUs) and be accessed remotely via the web interface allowing users to use the system simultaneously. For the training of the deep learning models, there is a queuing system. In addition, deep learning experts can easily plug in new neural network architectures and update the existing ones.
More Information:
- Grand Challenge/Workpackage(s): GC1 - WP4
- Responsible Research Leads: Aleksandra Pizurica
- Authors: Zeno Dhaene, Nina Zizakic, Shaoguang Huang
- Programming Language: Python
- Link to the repository
- Categories: Supervised learning
