autoStructN2V Training Process

Two-stage training with automatic structured noise detection and removal.

autoStructN2V training has three phases:

Stage 1: Initial N2V Training

Standard N2V training removes random noise. This produces an intermediate result where structured noise may still be visible.

Mask Extraction & Review

After Stage 1

• The system analyzes residual noise patterns

• A mask showing detected structure is generated

• You can review and adjust the mask before proceeding

• Options: Approve mask, regenerate with different settings, or skip Stage 2

Stage 2: Structure-Aware Training

A second model is trained using the noise mask. Masking follows the detected noise structure, teaching the network to specifically remove those patterns.

2.5D Mode Differences:

In 2.5D mode, the mask becomes 3-dimensional, capturing noise correlations across consecutive slices. The mask visualization shows three tabs (Z-1, Z center, Z+1) so you can inspect each slice's pattern before approving.

What to expect

• Total training time is roughly 2x single-stage N2V

• Stage 1 and Stage 2 can have different configurations

• The mask review step requires your input before Stage 2 begins

• Final results combine both stages for optimal noise removal