view article
Figure 1
Schematic drawing for the workflow of iterative particle training and picking. A small data set, usually 20–40 micrographs, is used for the workflow. The workflow comprises (a) an initialization step, (b) a training step followed by (c) a picking step. The process uses iterations of 2D class averages to improve particles for training a CNN for particle picking.

IUCrJ
Volume 7| Part 4| July 2020| Pages 719-727
ISSN: 2052-2525
https://doi.org/10.1107/S2052252520007241