Topaz negative stain data

Dear all,
is there anything to pay attention to when using topaz with neg. stain data (apart from setting neg. stain to true in the import job)?
Here is the result of my topaz picking after training, which is completely off (many particles not picked at all and also many wrong pickings (where there is no particle at all):

I’m sure there’s a way to make it work, but I’d ask yourself do you really need it? Topaz and other neural net pickers excel at difficult, low contrast situatons - NS is the opposite of that. Why not just use blob picker?

1 Like

Or manually pick a few dozen particles? You won’t need many particles to then get two or three classes for template picking. Although with NS data I’d back off the lowpass for templates/mics more than the default, say, 30-35A.

The thing is that the blob picker didnt work too well. However, I just remembered the last bit of the neg. stain tutorial section:

In some cases, CryoSPARC’s blob picker may not pick the highest-contrast particles in a negative stain dataset. If this occurs, picking may improve if the particle diameter is set to a lower value.

With the particle diameter set to 20-50 for my 150A particle, the blob picker works just fine