(HYPER-spectral image segmentation using deep neural NETworks)
Hyperspectral satellite imaging (HSI) has been gaining increasing attention due to the amount of information about the scanned region it can convey. Therefore, automated approaches for precise segmentation of such imagery are being rapidly developed, with deep-learning techniques constituting the mainstream.
We develop state of the art techniques which allow us to effectively analyse HSI with the use of deep-learning technology.
We work on cutting-edge algorithms to pre-process, augment, visualize, and precisely segment HSI, with the ultimate objective of helping practitioners better utilize such data
We design deep neural network models which can be seamlessly applied for segmenting multi- and hyperspectral satellite images. Our techniques have been thoroughly validated over a range of benchmark HSI datasets
We completed Hypernet within FP Space consortium. The project has been commissioned by the European Space Agency.