Argos Logo

Argos

Digital Image Correlation Software

See the crack, not the blur.

Standard DIC smooths displacement across a discontinuity. Argos2D puts the jump inside the correlation, so cracks, twins and slip bands come out sharp. Three methods in one GPU-native tool: classic subset DIC, Heaviside enrichment and neural correlation.

Argos2D field, neural correlation
Standard DIC field
Standard DIC
Argos2D

Effective strain (0 to 10 %) computed on the multi-modal dataset of J.C. Stinville, J.M. Hestroffer, M.A. Charpagne, A.T. Polonsky, M.P. Echlin, C.J. Torbet, V. Valle, K.E. Nygren, M.P. Miller, O. Klaas, A. Loghin, I.J. Beyerlein, T.M. Pollock, Scientific Data 9 (2022) 460. Argos2D result: neural correlation.

CUDA native Multi-GPU compatible Interactive visualization Crack & twin analysis Neural correlation

Method 2

Heaviside-Enriched DIC

Standard DIC assumes continuous displacement inside each subset, which smears out discontinuities across several pixels. Argos adds a Heaviside step function to the correlation model, allowing a sharp jump to be identified within each subset.

The subset shape function is augmented with a jump vector and a Heaviside function.

Used in Science Developed by V. Valle and co-workers at Institut PPRIME, Université de Poitiers. Method F. Bourdin et al., Measurements of plastic localization by heaviside-digital image correlation, Acta Materialia 157 (2018) 307–325. DOI Used in J.C. Stinville et al., On the origins of fatigue strength in crystalline metallic materials, Science 377 (2022) 1065–1071. DOI

φ(x) = x + u rigid body + ∇u · (xx0) first gradient Conventional DIC + u' · H(xx0) jump × Heaviside Heaviside-DIC

Method 3

Neural Correlation (xR-DIC)

xR-DIC is derived from the RAFT optical flow architecture, heavily modified for digital image correlation and trained on synthetic speckle images. It runs locally on your GPU: your images never leave your computer.

  • Per-pixel field: no subset size or spacing to tune
  • Fracture tab: cracks detected and measured directly on the neural field
  • Licensed option: the model and its engine are downloaded once, on first use
Effective strain computed with neural correlation (xR-DIC)
Effective strain computed with H-DIC
H-DIC
xR-DIC

Comparison of the effective strain between H-DIC and xR-DIC, on the dataset of J.C. Stinville et al., Scientific Data 9 (2022) 460.

A full GUI to set up correlations and explore the results.

Interactive Visualization

Load your images, draw the region of interest, run the correlation, then browse displacement, strain and fracture maps in the same window, with synchronized views, line profiles and per-field contrast.

Argos2D interface — displacement field with Heaviside enrichment

Dataset from A. Rouwane, D. Texier, J.-N. Périé, J.-E. Dufour, J.C. Stinville et al., Optics and Laser Technology, Vol. 177, 2024. Courtesy of Damien Texier.

Dedicated post-processing for cracks, twins and slip bands.

Fracture & Discontinuity Analysis

The fracture tab exposes the full enrichment output: crack opening displacement (intensity and direction), crack angle α, opening direction β, and a crystallographic decomposition into screw (longitudinal) and edge (transverse) components relative to the crack front.

Argos2D — Fracture tab showing crack opening fields

Dataset from A. Rouwane, D. Texier, J.-N. Périé, J.-E. Dufour, J.C. Stinville et al., Optics and Laser Technology, Vol. 177, 2024. Courtesy of Damien Texier.

Customizable derivative kernels and built-in criteria.

Strain Post-Processing

Compute the full strain tensor (εxx, εyy, εxy) from displacement fields using separable gradient operators.

  • Separable gradient kernels — Sobel, Scharr, centered finite differences, or custom
  • Custom expressions — built-in math parser for user-defined strain criteria
Argos2D — Strain tab with kernel selector and custom criterion editor

Dataset from A. Rouwane, D. Texier, J.-N. Périé, J.-E. Dufour, J.C. Stinville et al., Optics and Laser Technology, Vol. 177, 2024. Courtesy of Damien Texier.

Large-area analysis from tiled acquisitions.

Mosaic Stitching & Batch Mode

Assemble multi-tile acquisitions into a seamless mosaic before running DIC.

  • Automated grid layout — row/column organization with configurable overlap, handles missing tiles
  • Batch mode — process each tile pair independently without stitching, for high-throughput analysis
Argos2D — Mosaic stitching mode

Dataset from A. Rouwane, D. Texier, J.-N. Périé, J.-E. Dufour, J.C. Stinville et al., Optics and Laser Technology, Vol. 177, 2024. Courtesy of Damien Texier.

Core engine specifications

Under the Hood

Everything runs on your NVIDIA GPU, from pixel search to sub-pixel optimization and neural inference. Your images never leave your computer.

GPU Correlation CUDA-native correlation with multi-GPU support
Pixel Search Hierarchical coarse-to-fine + FFT-accelerated mode
Sub-pixel Optimization IC-GN (Inverse Compositional Gauss-Newton) with bilinear, bicubic, or B-spline cubic interpolation
Shape Functions Translation (2 DOF), affine (6 DOF), Heaviside-enriched (10 DOF), twinning mode (11 DOF)
Neural Correlation xR-DIC, a RAFT-derived recurrent network, tiled over the ROI for a dense per-pixel field
AI Inference ONNX Runtime on CUDA, fully local and offline, encrypted model

Get Argos2D

Free edition Free

The full correlation engine: displacement, strain and every correlation parameter.

Download free portable version
Full edition License key

Everything in the free edition, plus Heaviside enrichment, the fracture tab and mosaic stitching. Neural correlation is available as an option.

Request a license

Already have a key? Unzip the free edition and open Help → About Argos2D → Activate full edition. Argos2D validates your key online, then downloads and installs the full edition.

Feature Free Full
Correlation engine CUDA-native, multi-GPU, hierarchical and FFT pixel search included included
Displacement fields Sub-pixel IC-GN, bilinear / bicubic / B-spline interpolation included included
Strain post-processing Full tensor, selectable gradient kernels, custom strain criteria included included
Complete parameter set Subset, spacing, search area, NLC filter, ROI and mask, GPU selection included included
Image series and field export Reference plus any number of deformed images; every field exported to TIFF included included
Heaviside enrichment Discontinuity captured inside the subset; twinning mode not included included
Fracture tab Crack opening and direction, crack angle, screw / edge decomposition not included included
Mosaic stitching Multi-tile assembly, global position optimisation, batch mode not included included
Neural correlation (xR-DIC) Licensed option: dense per-pixel field, computed locally on the GPU not included option
In-app update notifications Downloading an update requires a license key not included included

System requirements

  • Windows 10/11 (64-bit)
  • NVIDIA GPU (minimum: Turing generation, e.g. GTX 16xx / RTX 20xx)

Questions, quotes, academic pricing: contact@argos-correlation.com

Coming Soon

Argos3D is currently in development.

Citing Argos2D

If Argos2D contributed to your published work, thank you for citing it. It helps a small independent tool reach the next laboratory.

Wantz, S. (2026). Argos2D (Version 1.2) [Computer software]. Sam'Strain. https://doi.org/10.5281/zenodo.22099000
BibTeX
@software{wantz_argos2d,
  author    = {Wantz, Samuel},
  title     = {{Argos2D}},
  version   = {1.2},
  year      = {2026},
  publisher = {Sam'Strain},
  doi       = {10.5281/zenodo.22099000},
  url       = {https://doi.org/10.5281/zenodo.22099000}
}

About the Author

Samuel Wantz, PhD

Founder & R&D Software Developer — Sam'Strain

Specialized in Digital Image Correlation (DIC) and Digital Volume Correlation (DVC) with expertise in experimental mechanics, image processing and X-ray Computed Tomography. PhD from Université de Poitiers, France, on metrological assessment and optimization of XR-CT and DVC protocols for mechanical testing.

Download the free portable version

Argos2D 1.3, Windows 10/11, 64-bit. Portable version.

Windows may warn you the first time you run Argos2D. Argos2D is a specialist tool with a small user base, so it has not yet built a download reputation. SHA-256: f9ff07b694fb6eb2f5de00fa87e83742a794f8fbc59f80357fc65f0b6b09f9b4