SIMUSCAN

SimuScan is a standalone software tool for the generation of realistic, labeled synthetic Atomic Force Microscopy (AFM) datasets.
It is designed to support the development, training, and benchmarking of deep-learning models for automated AFM analysis, including object detection and instance segmentation.

This release provides a precompiled executable version of SimuScan, enabling users to generate synthetic AFM datasets without requiring a Python installation or additional dependencies.

SimuScan simulates key physical and experimental aspects of AFM imaging, including:

• Tip convolution effects (single and double tips),

• Substrate tilt, curvature, and surface patterns,

• Realistic instrumental artifacts (noise, line defects, adhesion effects, drift, flattening),

• Parametric geometric and biological objects (e.g., bacteria, DNA, custom shapes),

• STL-based object rendering for complex 3D morphologies.

Dataset generation is fully controlled through a single YAML or JSON configuration file, allowing complete reproducibility.
For each run, SimuScan automatically stores the exact configuration used alongside the generated dataset.

The output follows the COCO dataset format, facilitating direct integration with modern computer-vision frameworks such as YOLO, Detectron2, and U-Net-based pipelines.

This software is intended for non-commercial academic and research use only.