CryoLens - simulated training data

The CryoLens training dataset consists of 14,366 synthetic tomograms containing 5.8 million particle instances across 103 protein structures ranging from 150 kDa to 2.7 MDa, with at least 46,000 instances of each particle type. The data was generated using copick-polnet at 10Å voxel spacing with a signal-to-noise ratio of 5.0, incorporating realistic cryo-ET artifacts including noise and missing wedge effects from a tilt-series range of -60° to +60°. The complete dataset totals approximately 12 terabytes, making it one of the largest synthetic training datasets available for cryo-electron tomography.