1. Current status and capabilities
2. Work in progress
3. Future plans, questions

ESRF uses a new tomography processing software: Nabu

Nabu is a versatile tomography reconstruction software.

• It offers a set of high performance building blocks
• On top of them, reconstruction pipelines are built for specific use cases:
  • Full-field absorption/phase contrast (now)
  • Diffraction tomography (work in progress)
  • Fluorescence tomography (in the future)

It primarily offers a Python API with a Cuda backend for performance-critical parts.

Support for absorption and phase contrast tomography comes with two pipelines.

Chunked Pipeline processes slabs of data volume.

It is fast and adapted to usual settings.

Grouped Pipeline processes groups of radios.

It is slower but addresses the shortcomings of the previous one.
Use case:

• Big datasets (eg. large number of projections)
• and
• Phase contrast with a large filter (CTF, Paganin with a large delta/beta)

Versatile

• Save checkpoints and resume the processing from a saved checkpoint
• Comes with a variety of tools:
  • Alignment (used by Bliss EBS-tomo)
  • Estimation (detector tilt, center of rotation)
• Python API makes it easy to prototype new features or even new pipelines

Fast

• All steps (except ID16A distortion correction) are available on GPU
• Full volume from 1200 projections (no HA): 2 mins
• Full volume from 7000 projections (no HA): 7 mins
• Full volume from 4400 projections (HA): 8 mins
• Reconstruction speed was limited by network data access - much better since GPFS deployment (still ~50% of the time in I/O)

• The current software processes the raw projections up to reconstructed volume.
  • Flat-field, CCD corrections, phase retrieval, de-striping, reconstruction, histogram
• For a list of available processing steps: silx.org/pub/nabu/doc/features.html

Still missing:

• Volume stitching
• Convert legacy tools/scripts from various beamlines
• Some of BM18 dedicated developments (see next slides)

• A new pipeline is being implemented for helical scans reconstruction.
• Need real data (from acquisition) for realistic tests

⚠️ This is not about larger-than-FOV correction.

Use case:

• do a quick scan of a large part of the sample
• view a low-resolution reconstruction
• select a region of interest
• enqueue a full-resolution reconstruction of this region

Ongoing work with Beamline Control Unit and people doing data visualisation

• Set of tools for interactive visualization (with user feed-back) of an ongoing (helical) scan
• Several technical questions raised (see next slides)

BM18 will require these additional developments on the ESRF side.

• Helical reconstruction
• Fan-beam/cone-beam reconstruction ?
• Interactive multi-resolution acquisition/reconstruction/visualization
• Compression ?
• Faster-than-FBP reconstruction methods
• Adapt Fraunhofer codes

Most of them are foreseen to benefit all ESRF tomography beamlines, however

• BM18 will require all these features "soon" to start smoothly
• Lack of resources on ESRF side (no one dedicated to BM18)

Compression:

• Compress radios on the fly ? If so, need a (very) fast encoder.
• The following algorithms are being considered:
  • Jpeg2000 (+ Part 15: HTJ2K)
  • ZFP
  • Jpeg-XL

Fast reconstruction methods (faster than FBP):

• Fourier-based
• Hierarchical FBP
• Log-polar

Adaptation of the Fraunhofer group codes

• "Code camp" at ESRF ?