Appion

Number of references to generate in an ab initio manner. This will depend on the amount of preferred orientation within your data. A good range to aim for could be ~100-1000 particles per reference, but because ML algorithms are computationally intense, this may not always be feasible, depending on the size of your dataset.