Inputs needed for preparation:

• density maps (one or more binned MRC or .rec files)
• trained model (.h5)

Output you will get:

Root directory

— Autopick:

— Autopick_[NAME]:

— data: folder storing data used to train the neural network

— test_x

— test_y

— train_x

— train_y

— metrics.png: image storing model metric

• Input Folder: Input folder for network predicting.
• Input Model: Input the trained model you obtained from network training.
• subtomo box size: The box size you used for predicting. The number should be divisible by 8.
• particle unit size: Unit size of your particles in pixels.
• min patch size: Specify the minimum size of a patch of an isolated area of lattice density that you want the model to predict particles for.
• y label size: Same number as the label size used to train your network.

ADVANCED

• tolerance: Specify the tolerance you want for bad particles.
• overlap size: Specify when particles can overlap

GUI for predicted particles after network training