Advanced Visualization Methods

High-dimensional data is hard to understand. But is it truly cursed? To help you develop better intuition for high-dimensional datasets, Nezar Abdennur and I built dtour, a visualization tool for smoothly navigating through data projections. → https://lnkd.in/ej9eFZJA Some dimensionality reduction methods find interesting angles but hide everything orthogonal to them. Other methods compress the whole manifold into 2D, which can muddle genuine structure with layout artifacts. Traversing multiple projections can help build a more holistic understanding of high-dim manifolds, and dtour makes such exploration effortless through one interface: you can switch fluidly between an overview gallery of projection “keyframes”, a guided cyclical tour along geodesic paths connecting keyframes, fine manual axis manipulation for user-driven excursions, and a wandering grand tour for serendipitous discovery. dtour is built around two types of workflows: 1. revealing structure in a single high-dimensional space through hyperdimensional tours, and  2. validating or comparing embeddings through sequential tours over a series of 2D projections. For examples: Dmitri Kobak and colleagues have shown that non-linear neighbor-embedding methods all lie on a spectrum that weighs attractive forces between neighbors against repulsive forces between all points. Smoothly scrubbing along that spectrum on Fashion MNIST makes it tangible which clusters in a UMAP-like layout are genuine signal and which only emerge under strong repulsion. You can also tour through the structure of a single high-dimensional dataset along its own spectral basis. For instance, on a 346K-cell single-cell surface-protein dataset from Florian Mair, a Laplacian Eigenmaps tour recovers known immunological hierarchy and gradually reveals cell phenotypes. On the validation side, we used dtour to inspect a 2D UMAP using a 8D PCA tour of a 276K-cell developing mouse brain from La Manno et al. showing which clusters are visible in the linear basis and which are exaggerated by the non-linear embedding. You can also use dtour to compare different embedding methods. For instance, a sequential tour through four embeddings of 3 million arXiv title+abstracts under SPECTER2, BGE-M3, Nomic Embed v2, and F2LLM-v2 8B surfaces where modern sentence embedding models agree on similarity and where they don't. Explore these yourself: https://dtour.dev dtour is optimized for performance: it runs smoothly for up to ~10M points and remains usable up to ~20M. To get there, we use WebGPU/GL, Web Workers, and OffscreenCanvas. dtour runs in any modern browser, ships as a Jupyter/Marimo widget for notebook workflows, and embeds in React apps. Preprint on arXiv: https://lnkd.in/e64r4-3M. And if dtour proves useful in any projection of your reality, please subscribe to a wonderful rest of your day.

Here's my 2024 LinkedIn Rewind, by Coauthor: 2024 proved that 3D Gaussian splatting isn't just another tech trend - it's transforming how we capture and understand the world around us. From real-time architectural visualization to autonomous vehicle training, we're seeing practical implementations I could only dream about a year ago. Through my "100 Days of Splats" project, I witnessed this technology evolve from research papers to real-world applications. We saw: → Large-scale scene reconstruction becoming practical → Real-time rendering reaching 60+ FPS → Integration with game engines and VFX pipelines → Adoption by major companies like Meta, Nvidia, and Varjo Three posts that captured pivotal developments: "VastGaussians - First Method for High-Quality Large Scene Reconstruction" Finally bridging the gap between research and AEC industry needs "This research is specifically tailored for visualization of large scenes such as commercial and industrial buildings, quarries, and landscapes." https://lnkd.in/gvgpqMNe "2D Gaussian Splatting vs Photogrammetry" The first radiance fields project producing truly accurate geometry "All in one pipeline I can generate a radiance field, textured mesh, and fly renderings - all in less than an hour" https://lnkd.in/geprBw6j "HybridNeRF Development" Pushing rendering speeds while maintaining quality "HybridNeRF looks better than 3DGS and can achieve over 60 FPS framerate" https://lnkd.in/gcqdE4iD Speaking at Geo Week showed me how hungry the industry is for practical applications of these technologies. We're no longer asking if Gaussian splatting will be useful - we're discovering new uses every day. 2025 will be about scaling practical applications - from AEC to geospatial to virtual production. The foundation is laid; now it's time to build. To everyone exploring and pushing the boundaries of 3D visualization - your experiments today are tomorrow's innovations. Keep building, keep sharing, keep pushing what's possible. #ComputerVision #3D #AI #GuassianSplatting #LinkedInRewind

Flying through 3D brain cell electron microscopy data in Unreal Engine with an Xbox controller in real-time: featuring open data from the IARPA MICrONS dataset. The video shows a selection of 250 mouse neurons, originally reconstructed in 3D from electron microscopy data by a large team of researchers from the MICrONS consortium. The rendering is a real-time capture of a 3D scene I set up in the game engine Unreal, using the 3D reconstruction data that has been generously made publicly available by the MICrONS consortium: https://lnkd.in/gDP5rpKv This cortical cubic millimeter data set spans a section of a mouse brain containing about 200,000 cells— so this 250-cell random subset represents less than 1% of the cell density in the original tissue. Each cell mesh is richly detailed, and even this sparse selection of cells is densely packed with data. The dataset has been a great resource for me to stress-test features in Unreal Engine for visualizing dense scientific spatial data. Unreal Engine recently released a new system for handling 3D scenes with dense geometry: Nanite. The idea behind Nanite is for geometry to automatically adapt in detail depending on how far it is from viewer, representing only the amount of detail that is actually perceptible from a given distance. As you get close, the representation becomes more detailed, and as you move away, it is simplified, freeing up resources for more critical parts of the scene. In this scene shown here, this method gives at least a 10-fold improvement in performance, transforming something that is barely workable to a fluid experience. These rendering approaches were originally developed for video games, which are constantly facing demands for increasing depth, scale, and immersion. But we’re facing similar challenges in scientific visualization: an explosion of depth and scale of scientific data and a need to find new ways to represent and engage with that data to make sense of it. ---------------------- References and acknowledgements: The data used for this visualization were produced by a consortium of labs led by members of the Allen Institute for Brain Science, Princeton University, and Baylor College of Medicine (the MICrONS Consortium). I was not involved in studies that generated the data. The dataset is described in the following publication: MICrONs Consortium, J. Alexander Bae, Chi Zhang, et al. Functional connectomics spanning multiple areas of mouse visual cortex. bioRxiv 2021.07.28.454025; doi: 10.1101/2021.07.28.454025 More information about the study, the Microns program, and these data can be found at: https://lnkd.in/gDP5rpKv This post was not sponsored or endorsed by the MICrONS Consortium, or any of its affiliated institutions or researchers. 

Advances in #multiplexed super-resolution microscopy are enabling researchers to understand the cellular complexity and protein function by visualising the intricate interactions and locations of numerous proteins #spatially within a cell. #KeyAdvancements driving this progress: #CyclicImmunofluorescence:These methods, already widely used in tissue imaging, are now being applied to subcellular mapping, providing multidimensional insights into protein localisation. #SuperResolutionMultiplexedImaging techniques: Methods like Exchange-PAINT, FLASH-PAINT, and SUM-PAINT are revolutionising our ability to image multiple proteins at the nanoscale level within cells. SUM-PAINT has even achieved 30-plex super-resolution imaging in neurons. #Value: Information-rich #AtlasProjects and #OpportunityToIntegrate with tissue imaging with nanoscale details of protein architecture in every cell. Key Links: https://lnkd.in/eqEgM7Nn: This research tagged 75% of the yeast proteome with GFP to image protein locations. https://lnkd.in/eHMGpk5D, https://lnkd.in/eHMGpk5D, https://lnkd.in/ePiHn8j4, https://lnkd.in/ePiHn8j4: These studies used CRISPR-based gene targeting for similar efforts in human cell lines. https://lnkd.in/eKsSDYw5: This paper introduced Exchange-PAINT, a pioneering method for multiplexed super-resolution imaging. https://lnkd.in/eHTrdTYv: This article details the development of FLASH-PAINT for multiplexed imaging in cells. https://lnkd.in/eSYkZK5t: This research presents SUM-PAINT, a method for highly multiplexed super-resolution imaging. Got value from this? Share 🔄 #SingleCell #TranslationalResearch #MultiOmics

🔥 Multi Omics Visualization Is Getting a Major Upgrade 📈 The study “MultiModalGraphics: an R package for graphical integration of multi omics datasets”, published in BMC Bioinformatics by Foziya Ahmed Mohammed, El Hadj Malick Fall, Kula Kekeba Tune, Rasha Hammamieh, Marti Jett and Seid Muhie, introduces a tool that turns complex biological data into visuals that actually make sense. 📌 Brings p values, q values, fold changes and feature counts directly into your plots so results speak for themselves 📌 Transforms heatmaps and volcano plots into intuitive, insight packed visuals for multi omics and multimodal datasets 📌 Lets you spot pathways, patterns and outliers in seconds instead of hours 📌 Connects with Bioconductor workflows so preprocessing, stats and visualization flow naturally 📌 Allows fast creation of multimodal figures that look polished and informative without extra coding 📌 Works across pan cancer datasets, mouse brain time series and multi tissue studies to reveal hidden biological signals 📌 Makes quantitative annotation accessible to everyone, even if you are not a visualization expert 📢 Join the Conversation 📢 Share your ideas, methods, and tools in the comments! 👇 💬 👉 Follow my blog for more https://lnkd.in/dvppv8uc

#OA Review article Visualizing the Future: Recent Progress and Challenges on Advanced Imaging Characterization for All-Solid-State Batteries | ACS Energy Letters https://lnkd.in/ggZUxE4B This review highlights recent progress in cutting-edge visualization techniques, including neutron imaging, X-ray tomography, focused ion beam scanning electron microscopy, and cryogenic electron microscopy, which reveal microstructural and chemical changes in ASSBs at scales from the atomic to the macroscopic level. It focuses on the elusive failure behaviors at lithium anodes, composite cathodes, solid-state electrolytes, and beyond. 

Visualizing chemically complex alloy systems (e.g. high entropy alloys) is quite challenging. Yet, without reliable visualization tools, it is impossible for us humans to build an intuitive understanding of the multi-dimensional materials spaces we are trying to explore and exploit in order to find the next materials that help us solve some of the most pressing challenges. Today, my students Brent Vela and Trevor Hastings just uploaded a manuscript to the ArXiv where they describe a solution to this problem through the use of Uniform Manifold Approximation and Projection (UMAPs). UMAP is a really cool dimensional reduction technique that preserves both local and global structures in multi-dimensional datasets. We have been using them for a while in many of our papers and presentations, but this manuscript provides a more in-depth discussion on best practices to the visualization of large dimensional alloy spaces. Moreover, Brent and Trevor have created a Code Ocean repository to disseminate the toolkit. Paper is here: https://lnkd.in/dkGSzGE7 To test the toolkit, go to: https://lnkd.in/dThkAfdJ