The Holoflow project

Real-time digital holography, from acquisition to insight.

Holoflow is an open-source C++ and CUDA stack for describing, running, and exploring high-throughput scientific imaging pipelines. Holovibes brings the stack into the lab through an interactive Qt application.

Explore the project View on GitHub
Laser Doppler angiography processed in real time.12

Experiments and engineering

What we build with it

One stack connects the processing model, the GPU runtime, and the measurements seen in the lab.

Holoflow graph connecting acquisition, GPU processing tasks, and display output

A graph backend for scientific pipelines

Describe the computation as connected tasks. Holoflow compiles the graph, manages tensor lifetimes and memory, then schedules CPU and GPU work while the algorithm remains readable.1

Understand the runtime
Interferometric input
Doppler angiography

Laser Doppler holography

Transform high-rate interferometric acquisitions into live views of retinal blood flow, with reconstruction and visualization performed as the data arrives.12

Reproduce the experiment
Interferometric input
Doppler OCT
OCT footage coming soon

Doppler OCT

Use the same modular processing foundation to explore depth-resolved Doppler measurements and evolve experimental pipelines without rebuilding the application around each method.

Open the application guide
Interferometric input
Wavefront analysis
Wavefront footage coming soon

Wavefront analysis

Inspect wavefront quality as the experiment runs, including Shack–Hartmann measurements and Zernike modes used to understand and correct optical aberrations.

Open the application guide
Interferometric input
Tomographic diffractive microscopy
TDM footage coming soon

Tomographic diffractive microscopy

Combine off-axis holography, aberration correction, and phase unwrapping to reconstruct three-dimensional sample volumes from tomographic acquisitions.

Open the application guide

One project, several layers

Choose the level you need

Start with the application or work directly with the runtime and its infrastructure libraries.

Desktop application

Holovibes

Acquire, reconstruct, analyze, and visualize holographic data interactively.

Explore Holovibes Execution runtime

Holoflow

Compile declarative processing graphs into predictable CPU and GPU execution.

Explore Holoflow
GPU infrastructure

Curaii

RAII wrappers for CUDA resources and companion libraries.

Explore Curaii Acquisition format

Holofile

High-throughput reading and writing of holographic recordings.

Explore Holofile

Supporting libraries: holotask for reusable operators, holonp for numerical primitives, and holoflow_event for runtime communication.

Built by

Physics and engineering, developed together

Michael Atlan

Michael Atlan

Scientific lead · Tenured Researcher at CNRS

Michael defines the scientific direction across digital holography, Doppler imaging, optical coherence tomography, and ophthalmic experiments.

Research profile
Jules Guillou

Jules Guillou

Creator and lead developer

Jules designs and implements the software stack, from the Holoflow execution model and CUDA processing to the Holovibes application.

GitHub profile

Where to begin

Enter through the experiment or the engine.

Explore Holovibes Understand Holoflow

  1. L. Puyo, M. Paques, M. Fink, J.-A. Sahel, and M. Atlan, “In vivo laser Doppler holography of the human retina,” Biomedical Optics Express, vol. 9, no. 9, pp. 4113–4129, 2018. https://doi.org/10.1364/BOE.9.004113. See also the site-wide reference. ↩

  2. M. Atlan, Doppler Holography Measurements of the Eye Fundus in a Volunteer – May 27, 2025 [Data set]. Zenodo, 2025. https://doi.org/10.5281/zenodo.16761111. Licensed under CC BY 4.0. See also the site-wide reference. ↩