Python API + Batch Automation: EikoTwin 2026.1 is turning DIC calibration from a manual task into an assembly-line process!
Release time:
2026-10-09 10:20
Source:
DIC can already provide full-field strain data, but between “providing” that data and “putting it to use” lies an entire cycle of repetitive tasks: importing data frame by frame, selecting points one by one, running simulations for each case, and compiling reports manually. The 2026.1 release of EikoTwin brings the Python API and batch processing front and center—it aims to optimize not measurement accuracy, but the throughput of the downstream workflow following measurement.
01 Application of Python API and Batch Automation in Strain Measurement
PYTHON API & BATCH AUTOMATION FOR DIC
Labor-intensive operations represent an invisible ceiling for the implementation of DIC.
A review on damage identification in engineering materials puts it bluntly: traditional imaging and DIC can provide valuable information on deformation and crack evolution, but “often require substantial manual effort and suffer from accuracy limitations under complex loading conditions.” [1] The problem rarely lies with the algorithms themselves, but rather with the measurements: a single formal experiment can easily involve hundreds or even thousands of frames and dozens of operating conditions, with frame-by-frame processing and case-by-case modeling relying almost entirely on manual effort.
What’s even more problematic is that the more manual steps there are, the harder it becomes to ensure reproducibility. The field of image processing has long had a well-established automated pipeline: acquisition → preprocessing → segmentation → feature extraction → detection → post-processing, with each stage aiming to “reduce human intervention and improve consistency.” [2] . The data flow of DIC is essentially no different; what’s lacking is only an engineered encapsulation that replicates this pipeline as-is.
The True Value of APIs and Batch Processing: Turning “Operations” into “Scripts”
When measurement software provides a Python API, the entire workflow can be scripted: batch‑reading image sequences, batch‑defining regions of interest (ROIs), and batch‑exporting displacement/strain fields, which can then be fed to downstream scripts for statistical analysis or inversion. The open‑source community has already demonstrated such an approach—a DIC application for humidity‑induced deformation in oil paintings that even eliminates the need for artificial speckle generation, instead using the canvas’s natural texture for correlation, and includes comprehensive metrological evaluation. [3] ; Further advances in satellite imagery-related fields employ deep-learning autoencoders to automatically remove relevant noise, enabling the extraction of displacement fields without manual intervention. [4] 。
“The significance of automating measurement results lies not in saving a few clicks, but in ensuring that every click can be reproduced.”
Another layer of value in batch processing lies in “data volume.” When full-field displacements can be exported automatically and in batches, they can, in turn, be used to train models: a fatigue crack growth dataset was automatically collected using a robotic test rig combined with a commercial DIC system. 8,794 displacement fields , and then, through standardized interpolation and augmentation, it is expanded into 70,352 supervision samples and directly marks the coordinates of the crack tip and the stress intensity factor. [5] 。

Figure 1: DIC displacement field dataset of fatigue cracks automatically acquired by the robotic test rig (adapted from Ref. [5])
02 EikoTwin Evaluation of Strain and Displacement Measurement Performance
STRAIN & DISPLACEMENT MEASUREMENT PERFORMANCE
Data quality determines how far automation can go.
Automation is not about processing bad data faster. To enable scripts to replace humans, spatial integrity across the experimental domain becomes critical. In one composite material study, a single DIC strain map at a nominal strain of 0.2% was used to predict the entire shear stress–strain curve, with an average error over five-fold cross-validation. R² = 0.926 ± 0.022, RMSE = 6.37 ± 1.14 MPa ; Data augmentation and Dropout reduce RMSE by another ~40% [6] Such “low-input, full-output” models are extremely sensitive to the completeness of the input field’s boundaries and the consistency of its coordinates.

Figure 2: Prediction of the shear stress–strain curve for a composite material from a single DIC speckle field (adapted from Ref. [6])
EikoTwin DIC The key advantage lies here: the DIC algorithm is built on a finite-element mesh, with displacements and strains directly defined at the nodes, allowing for straightforward comparison with simulation results in the same coordinate system. This eliminates the intermediate steps of projection and interpolation—processes that are both error-prone and difficult to script. For batch processing, this means that each case generates output with a consistent structure and uniform fields, enabling scripts to iterate reliably without requiring separate analytical logic for every specimen.
Evidence-based results: Measured data from rotating components using non-contact measurement
The effects can be quantified. 3D DIC was used to measure the full-field displacement and strain of a rotating propeller, and compared with FEM in 4110, 6045, 6940 rpm Point-by-point comparison at three speeds shows good agreement in the selected region. [7] . In another multi-point dynamic test, the first bending frequencies measured by 3D DIC and the accelerometer were respectively 49.1 Hz and 48.7 Hz, with a deviation of only 0.82%, and the minimum MAC reaches 0.91 —And this comparison is likewise generated in bulk through standardized processes.
03 Solution Analysis
SOLUTION BREAKDOWN
Parameter identification, shortened from “several days to several weeks” to a single forward pass.
Traditional parameter identification typically focuses only on global curves such as force–displacement, reserving full-field strain for post‑hoc verification—a trade‑off that requires repeated finite element iterations; a single calibration run on a high‑performance computing system can take from days to weeks. Surrogate models are rewriting this workflow: a 3D graph neural network trained on FE full‑field data can reproduce the complete evolution of stress, strain, and force–displacement, replacing iterative simulations with a single forward pass. [8] ; and the conventional approach of converting an FE mesh into an image/voxel leads to either loss of accuracy or memory bloat—this precisely demonstrates that, The grid’s native data format is inherently efficient. 。

Figure 3: Comparison of the graph neural network surrogate model with FEM full-field responses (reproduced from Ref. [8])
Automation is also making inroads into the “identification” stage. For modal parameter identification of offshore wind turbine support structures, an improved clustering method automatically identifies the first three modal frequencies, with deviations from reference values of only 0.0%, 0.30%, 0.18% [9] ; On the materials side, U-Net reconstructs the full-field temperature distribution solely from DIC strain fields, R² = 0.985 , making it possible to obtain thermal fields without an infrared camera [10] 。

Figure 4: Modal vibration shapes for the automatic identification of offshore wind turbine support structures (adapted from reference [9])
Putting these together, EikoTwin 2026.1 The goal is to deliver an end-to-end pipeline: camera capture → native mesh DIC → batch export via Python API → scripted parameter identification → automated report generation. The user shifts from being an “operator” back to being a “rule-definer.”
Turning DIC from a “manual process” into an “assembly line” hinges not on faster algorithms, but on reducing and making reproducible the amount of manual intervention required.
References
[1] SADEGHIAN, PALEVICIUS, SABLINSKAS, et al. From Pixels to Predictions: Integrating Machine Learning and Digital Image Correlation for Damage Identification in Engineering Materials[J]. Materials, 2025, 19(1): 77. DOI: 10.3390/ma19010077.
[2] PATEL, DESAI. An Automated Image-Based System for Detection of Surface Defects Using Digital Image Processing[J]. International Journal of Computer Applications, 2026. DOI: 10.5120/ijcabe66bf09854b.
[3] LIAO, GIBSON, GRAU-BOVÉ, et al. Open-source digital image correlation application for monitoring humidity-induced deformation in canvas paintings with metrological evaluation[J]. npj Heritage Science, 2026. DOI: 10.1038/s40494-026-02415-9.
[4] LI, WU, WANG, et al. Reconstructing Horizontal Displacement Through Deep Learning in Multiple-Pairwise Satellite Image Correlation[J]. Remote Sensing, 2026, 18(5): 704. DOI: 10.3390/rs18050704.
[5] MELCHING, DÖMLING, PAYSAN, et al. Annotated digital image correlation displacement fields from fatigue crack growth experiments on aluminium alloys[J]. Scientific Data, 2026. DOI: 10.1038/s41597-026-07722-1.
[6] RUAN, YAO, ZHAO, et al. Individualized Prediction of In-Plane Shear Stress–Strain Curves for Composites Using Early-Stage Digital Image Correlation Strain Fields[J]. Materials, 2026, 19(12): 2609. DOI: 10.3390/ma19122609.
[7] PAZUR, SPYCHAŁA, MACIOROWSKI, et al. Full-Field Displacement and Strain Measurement of a Rotating Propeller Using 3D Digital Image Correlation and FEM Analysis[J]. Applied Sciences, 2026, 16(15): 7754. DOI: 10.3390/app16157754.
[8] MODAD, MAKKI, CHEHADE, et al. A graph neural network surrogate for 3D finite element modeling: Accelerated full-field parameter identification in aluminum alloy 6DR1[J]. Journal of Materials Science: Materials Theory, 2026, 10: 5. DOI: 10.1186/s41313-026-00075-7.
[9] YANG, LIANG, ZHU, et al. Automatic Modal Parameter Identification for Offshore Wind Turbines Using Modified Clustering-Based Methodology[J]. Sensors, 2026, 26(8): 2536. DOI: 10.3390/s26082536.
[10] GREBO, NOVAK, PANIĆ, et al. A Deep Learning Framework for Full-Field Thermal Field Distribution Prediction from Digital Image Correlation Strain Measurements[J]. Applied Sciences, 2026, 16(1): 460. DOI: 10.3390/app16010460.
EikoTwin DIC,Automation of DIC Measurement Batch Processing,FE Simulation Comparison and Validation,Python API + Batch Automation,Non-contact Full-field Strain Measurement