All research
Research method Under review

RAFTcorr

A learning-accelerated DIC method for dense, sub-pixel displacement measurement, characterized against rigid motion, large deformation, noise, and complex specimens.

DICExperimental MechanicsDense MeasurementMachine Learning
Code & implementation Paper

RAFTcorr is the first fully open-source RAFT-based Digital Image Correlation framework, bridging deep learning and experimental mechanics for dense displacement measurement.

Key Features:

  • Physically realistic training-data generation pipeline
  • Complete training scripts and pre-trained model weights
  • User-friendly GUI that eliminates manual parameter tuning
  • Sub-pixel accuracy across rigid-body translation, rotation, large deformation, and complex metamaterial geometries

Status: Under review — preprint: RAFTcorr: A Deep Learning Digital Image Correlation Framework with Operating-Boundary Characterization

Demo Results

Aluminum Plate with Hole

Displacement field
Von Mises strain field

Cavitation Flow

Displacement field
Velocity magnitude & streamline

Foam Fracture

Displacement field during foam fracture