Joshua Dare-Cullen

I am a PhD researcher focused on advancing methods for hyperspectral image compression, with a particular interest in neural compression techniques. My work explores the intersection of deep learning and remote sensing, aiming to design novel architectures that efficiently process and compress the rich spectral and spatial information in hyperspectral data. I am currently developing approaches that combine spectral modeling using state-space models like Mamba with innovative preprocessing and attention mechanisms to enhance feature extraction and compression performance. Additionally, I am exploring the integration of entropy modeling for accurate compress-decompress workflows, targeting applications in earth observation and geospatial intelligence. My research is supported by an industrial partnership with Ordnance Survey, and I am keen to bridge the gap between academic research and real-world applications.