Seminars and lectures at ESRF
Here is the list of seminars and lectures held in the context of the tomo-bridge.
-
Presenter: Joan Rue Queralt from: EPFL - Center for Imaging
Title: "Pyxu: Simplifying Advanced X-ray Imaging"
Abstract: Pyxu (https://pyxu-org.github.io/) is a new, user-friendly software tool designed to make advanced X-ray imaging techniques accessible to a broader range of professionals. It eliminates the need for deep mathematical or programming expertise, allowing users to easily develop and implement sophisticated imaging methods. This open-source Python library offers a variety of ready-to-use imaging tools and functions, which can be combined like building blocks to create custom solutions. Pyxu is efficient for large-scale tasks and incorporates the latest in image reconstruction, including AI-based denoisers, reliability assessments, and automatic adjustments for optimal performance. I will showcase Pyxu's capabilities through an example in computational tomography, highlighting its ability to produce high-quality images and adapt to complex imaging setups.
Room: zoom
Date and time: Thu Dec 7th 2023, at 14h00
Slides: link
Recording: link
-
Presenter: Jakob Sauer Jørgensen from: DTU
Title: "Tomographic reconstruction and uncertainty quantification with CUQIpy and the Core Imaging Library"
Abstract: Scientific discoveries in imaging increasingly depend on computational processing of the raw data, and the widespread uptake of advanced computational methods relies on powerful, yet easy-to-use, open-source software libraries. In this talk, we give an overview of two open-source Python libraries: The Core Imaging Library (CIL), available from https://ccpi.ac.uk/cil/, is a versatile framework for processing and especially optimization-based reconstruction of many types of CT data, including parallel and cone-beam, high-noise, dynamic and spectral CT, where data quality is low or in non-standard configurations.CUQIpy (pronounced “cookie pie”), available from https://cuqi-dtu.github.io/CUQIpy/, is a new library for Bayesian inversion and computational uncertainty quantification for inverse problems including CT reconstruction. We show example applications for both libraries and discuss future directions.
Room: zoom
Date and time: Thu Jun 15th 2023, at 14h00
Recording: link
-
Presenter: Ozan Öktem from: KTH - Department of mathematics
Title: "Microlocal analysis and deep learning for limited angle tomography."
Abstract:
Room: zoom
Date and time: Thu Feb 23rd 2023, at 14h00
Recording: link
-
Presenter: Patrik Vagovic from: XFEL
Title: "MHz X-ray Microscopy at European XFEL"
Abstract: The European XFEL [1], provides unique opportunities for the characterisation of stochastic dynamics occurring in various systems, either naturally or in response to stimulation by an external force for example induced by lasers or mechanical forces. Properties of EuXFEL probe allow to obtain ultrashort (fs scale) X-ray snapshots with up to 4.5 million such snapshots per second [2]. This enables one to film individual realisations of fast stochastic processes in slow smooth motion at micro-meter spatial resolution capturing object velocities up to ~km/s scale. We will show recent progress and examples of stochastic dynamics imaged at SPB/SFX instrument of EuXFEL using MHz radiography. For complex dynamic structures access to 3D information is necessary. However, at MHz sampling rates is not possible by conventional approaches such as X-ray tomography. To be able to obtain volumetric information at such high frame rates (MHz) only possibility is to create set of probes rotated angularly around sample. Each XFEL pulse then provide simultaneous angular views which can provide 3D information. We will present preliminary results from first feasibility study of novel multi-projection imaging system for XFEL’s supported by EIC-Pathfinder MHz-Tomoscopy project [3]. This development will open up new possibilities to inspect dynamic matter in slow motion in 3D using hard X-rays 24keV -30keV. References: [1] W. Decking et al., Nature Photonics 14, 391–397 (2020), [2] P. Vagovič, et al., Optica 6, 1106-1109 (2019), [3] EIC-Pathfinder MHz-Tomoscopy, Nov. 2021, Coordinator: Patrik Vagovič, CFEL, DESY.
Room: zoom
Date and time: Thu Jan 13th 2022, at 14h00
-
Presenter: Allard Hendriksen from: CWI, Amsterdam, The Netherlands
Title: "Self-supervised deep denoising for synchrotron tomography"
Abstract: In synchrotron tomography, experimental constraints can impose dose and time constraints that leads to noise that carries over into the reconstructed images. Convolutional neural networks (CNNs) have rapidly gained popularity as a powerful tool for removing noise from reconstructed images. However, training CNNs typically requires collecting a dataset of paired noisy and high-quality measurements, which is a major obstacle to their use in practice. To circumvent this problem, we have proposed a method for training deep convolutional neural networks to denoise tomographic reconstructions that does not require any noise-free data. We present results on challenging dynamic micro-tomography and X-ray diffraction computed tomography datasets. In addition, we discuss challenges and provide an outlook on future work.
Room: zoom
Date and time: Thu Jul 1st 2021, at 14h00
Recording: link
-
Presenter: Willem Jan Palenstijn from: CWI, Amsterdam, The Netherlands
Title: "Software for real-time tomographic reconstruction: ASTRA, Pleiades and Recast3D"
Abstract: We will discuss a number of ongoing software projects at the Computational Imaging group at CWI, Amsterdam around real-time tomographic reconstruction: ASTRA for GPU acceleration, Pleiades for distributed 3D reconstruction, and Recast3D for real-time reconstruction, visualization and processing of arbitrarily oriented slices. We will look at recent work, and also briefly at future plans.
Room: LOB 1-45
Date and time: Wed Feb 5th 2020, at 09h00
-
Presenter: Joost Batenburg, Willem Jan Palenstijn, Sophia Bethany Coban, Felix Lucka from: CWI, Amsterdam, The Netherlands
Title: "Developing the next generation tomography solutions"
Abstract: Next generation synchrotron tomography research and development imposes many new challenges that go beyond the state-of-the-art in available algorithms and software. Strongly increasing data rates and sizes make us revisit the traditional workflow of 'first store, analyze later', while at the same time the ability to perform imaging at nm-range resolutions make the computational needs in processing the data more complex. While standard workflows in processing synchrotron tomography data are still mostly based on 2D slices, new tomographic techniques often require a fully-3D approach which is more difficult to parallelize, imposing high computational requirements. In this seminar, we present the key efforts of the CWI Computational Imaging group to deal simultaneously with the challenges of multi-modality, high-throughput, big data, and limited data. A special emphasis will be on the role of recent machine learning approaches, and the challenges in making them applicable to large-scale 3D image data.
Room: LOB 1-45
Date and time: Wed Jul 10th 2019, at 14h00
Slides: link