cryoSENSE
Published:
Compressive Sensing with Generative Priors on the Protein Cryo-EM Image Manifold
Diffusion · Inverse Problems · Cryo-EM Imaging — IEEE/CVF CVPR 2026
Project website Paper (arXiv) CVPR paper (PDF)
Cryo-electron microscopy (cryo-EM) enables atomic-resolution visualization of biomolecules, but data acquisition is expensive and low-throughput. cryoSENSE leverages the low-dimensional manifold of protein cryo-EM images, combining compressive sensing with sparse and generative (diffusion) priors to reconstruct high-resolution images from far fewer measurements — increasing acquisition throughput by up to 2.5×.
