topaz.protocols.protocol_topaz_training module

class topaz.protocols.protocol_topaz_training.TopazProtTraining(**args)

Bases: pwem.protocols.protocol_particles_picking.ProtParticlePicking, topaz.protocols.protocol_base.ProtTopazBase

Train and save a Topaz model

ADD_MODEL_TRAIN_MODEL = 1

ADD_MODEL_TRAIN_NEW = 0

ADD_MODEL_TRAIN_TYPES = ['New', 'TopazModel']

convertInputStep(inputCoordinates, scale, kfold)

Converts a set of coordinates to box files and binaries to mrc if needed. It generates 2 folders 1 for the box files and another for the mrc files.

createOutputStep()

Register the output model.

denoiseStep()

getNNModelFn()

Returns the model fn (or type) as expected from topaz software

getPickingFileName(micList, key)

preprocessStep()

Downsamples the micrographs with a factor determined by the scale parameter and normalize them with the per-micrograph scaled Gaussian mixture model

trainingStep(radius, enc, numEpochs, modelFit, method, numParts, extra)

Train the model with the provided parameters and the previously preprocessed micrograph images and the provided input coordinates.