Diamond Light Source’s Post

Transfusion: Predict the Next Token and Diffuse Images with One Multi-Modal Model Meta's new paper introduces Transfusion, a novel approach for training a model that can seamlessly handle discrete and continuous modalities. The model is pre-trained on a balanced mix of 50% text and 50% image data, using distinct objectives for each modality: next token prediction for text and diffusion for images. When compared to Chameleon, Transfusion demonstrates superior scalability across various modality combinations. It delivers better FID and CLIP scores for text-to-image generation while using less than one-third of the compute and achieves comparable perplexity with only 50%- 60% of Chameleon's FLOPs. Remarkably, the 7B Transfusion model surpasses DALL-E 2 and SDXL in performance and matches Llama 1’s results on text benchmarks. Paper: https://lnkd.in/dYp2Tni2

!!! Explore Spectroscopy Data with Avizo Software !!! Join this webinar to learn how to: * Visualize spectroscopy data in 3 dimensions * Correlate spectroscopy and imaging data for an in-depth material understanding * Facilitate segmentation leveraging on spectroscopy information https://lnkd.in/d8SyuFex

Do you want to be part of the CHARISMA Raman School 2024? 🚀 This edition focuses on the relevance, impact, needs, state-of-the-art of harmonisation in different techniques and fields of application. Topics that will be covered include: • Data processing and analysis • Raman harmonisation and standards • In vivo, in situ and operando Raman characterisation • Applications of Raman spectroscopy and market opportunities • Raman data repositories and FAIR data 📢 Register now: https://lnkd.in/d9xdvke7 #ramanspectroscopy #ramantechniques

Our dissoLab webinar is coming up soon. Please join us if interested to discover the future of dissolution! https://lnkd.in/ggnPuDiA In addition to introducing the platform, we will cover the following learning objectives: 1. Advantages of image based dissolution modeling over data dependent regression models 2. Use of digital models for rapid dissolution assessment 3. How particle morphology plays a vital role in dissolution

Did you know that X-ray crystallography, a special X-ray technique, was used to discover the double-helix shape of DNA in the early 1950s? For more fun facts, download our digital toolkit. https://lnkd.in/gYmANEiN

It's that time of the week again – #UnderTheMicroscopeMonday 🔎🔬! We are thrilled to present an incredible data image that provides an in-depth cross-sectional analysis of a standard business card. To capture the data image, we employed the XploreCompact EDS detector in conjunction with the AZtecLive real-time chemical imaging tool. This combination allowed us to effectively analyse and visualise the chemical composition of the sample in real-time. #CrossSectionAnalysis #EnergyDispersiveXraySpectroscopy #EnergyDispersiveSpectroscopy #ChemicalImaging

It's that time of the week again – #UnderTheMicroscopeMonday 🔎🔬! We are thrilled to present an incredible data image that provides an in-depth cross-sectional analysis of a standard business card. To capture the data image, we employed the XploreCompact EDS detector in conjunction with the AZtecLive real-time chemical imaging tool. This combination allowed us to effectively analyse and visualise the chemical composition of the sample in real-time. #CrossSectionAnalysis #EnergyDispersiveXraySpectroscopy #EnergyDispersiveSpectroscopy #ChemicalImaging

It's that time of the week again – #UnderTheMicroscopeMonday 🔎🔬! We are thrilled to present an incredible data image that provides an in-depth cross-sectional analysis of a standard business card. To capture the data image, we employed the XploreCompact EDS detector in conjunction with the AZtecLive real-time chemical imaging tool. This combination allowed us to effectively analyse and visualise the chemical composition of the sample in real-time. #CrossSectionAnalysis #EnergyDispersiveXraySpectroscopy #EnergyDispersiveSpectroscopy #ChemicalImaging

Paper : Restoration algorithms and system performance evaluation for active imagers Abstract : This paper deals with two fields related to active imaging system. First, we begin to explore image processing algorithms to restore the artefacts like speckle, scintillation and image dancing caused by atmospheric turbulence. Next, we examine how to evaluate the performance of this kind of systems. To do this task, we propose a modified version of the german TRM3 metric which permits to get MTF-like measures. We use the database acquired during NATO-TG40 field trials to make our tests. 

New approach for motion detection and motion tracking has been presented on ICST conference. “Video pre-motion detection by fragment processing”. Proposed approach will not answer the question of which object is moving, so there is no talk about classifying moving objects. The combination of fragment analysis and SVD allows finding fragments of the frame that could serve as a region of interest, and to which a more reliable algorithm with analysis only the necessary information of a much smaller size compared to the initial data could be applied. In this study a Ky Fan norm matching based SVD-covariance descriptor for motion detection and motion tracking was proposed. Methodology describes the SVD generation of the region covariance as the tracking feature. https://lnkd.in/d2iqTyhj