iiinova — where your digital self keeps evolving

Reposted by Constantin Pape

Davi Bock @dddavi.bsky.social · 24/09/2026

Now I can say: last week we received Notice of Award from NSF on a Midscale Research Infrastructure Design proposal for a National Connectomics and Ultrastructure Volume Observatory (NCUVO) www.uvm.edu/uvmnews/news...

uvm.edu

UVM Launches $19 Million National Observatory to Unlock…

| The University of Vermont

137317

Constantin Pape @cppape.bsky.social · 28/08/2026

It seems more reasonable to just commit the code and results at this point and let a coding agent check if the code matches claims. Any similar experiences?

011

Constantin Pape @cppape.bsky.social · 28/08/2026

Leading to well sounding, but extremely dense and jargon-heavy prose. Combined in many cases with very bad experiment design (a lot of baselines that don't fit --- asked coding agent to just compare to everything that seems relevant?).

110

Constantin Pape @cppape.bsky.social · 28/08/2026

I feel like giving up on reviewing for most biomedical imaging focused CS journals. Over the last year almost all papers I got where basically AI generated (probably the underlying experiments are real, but the descriptions seemed to be filtered through several AI rewrites)

130

Constantin Pape @cppape.bsky.social · 16/07/2026

This research will help us to build better tools for interactive microscopy analysis --- stay tuned for more! For more details: t.co/94KnNtCXk7 Thanks to Caro for driving this work and all other co-authors!

t.co

https://arxiv.org/abs/2603.19802

040

Constantin Pape @cppape.bsky.social · 16/07/2026

We compare different foundation models (SAM and DINO variants) with classical features in ilastik and fully supervised learning. The FM features outperform classical approaches. Attentive probing yields better performance than RF, but is too slow for interactive use. See example results in the fig.

Results for object segmentation based on foundation model (FM) features with a random forest, compared with supervised learning baselines and classical features. Experiments over five different datasets covering cell classification in spatial proteomics (CRC, HBM), high-content macroscopic imaging for animal phenotyping (Planri), nucleus classification in histopathology (PanNuke), and cell line classification in phase-contrast microscopy (LIVECell). The extra plot repots runtimes for model training and inference.

120

Constantin Pape @cppape.bsky.social · 16/07/2026

We use embeddings from foundation models as features, either for a RF or different attentive probing variants. The features are averaged over masks for object classification. See the image for a method overview.

120

Constantin Pape @cppape.bsky.social · 16/07/2026

Segmenting individual cells in microscopy is much easier these days thanks to foundation models. Can we use these models for other tasks, such as cell classification? We investigate in our latest work, finding big improvements for object and pixel classification compared to classical approaches.

Evaluation of different methods for pixel classification (top) and object classification (bottom) on the LIVECell dataset. Dark green bars show the F1-Score, which measures the classification / segmentation quality (higher is better), light green bars show the runtimes. Five different settings are compared for each task: Ilastik features + random forest (RF), microSAM embeddings + RF, SAM2 embeddings + RF, and uSAM, SAM2 + attentive probing (DeAP and ObAP). microSAM feature perform best for RF based methods, attentive probing outperforms RF based approaches, but at a much higher runtime.

1197

Constantin Pape @cppape.bsky.social · 12/07/2026

In summary: great to have a second useful coding agent on par or even better than Claude Code + Opus 4.8, BUT I would prefer a Fable that doesn't reject any reasonable request ...

160

Constantin Pape @cppape.bsky.social · 12/07/2026

probably explaining some of the unhinged behavior. Interesting side-effect: generating reports on optimization strategies with Codex / 5.6 enables getting past Fable safeguards (which is still a superior model); it accepts validating and improving upon it while it would reject the work from scratch.

110

Constantin Pape @cppape.bsky.social · 12/07/2026

Or: tries to edit files by executing hex code python commands when apply-patch fails. 5.6-Terra: not useful, watching it to struggle compile is painful ... Overall impression: 5.6-Sol > Opus 4.8 (it did manage some optimizations Opus couldn't pull off). Codex seems to be a worse harness though,

100

Constantin Pape @cppape.bsky.social · 12/07/2026

Initial thoughts on 5.6-Sol: actually makes codex useful (was not helpful for C++ with 5.5 for me at all). Still unhinged compared to Claude (large performance critical refactors without careful benchmarking when not being explicitly told -> lead to major regressions, Opus would not do this)

120

Reposted by Constantin Pape

Anwai Archit @anwaiarchit.bsky.social · 12/07/2026

Hi all, Felt like a good opportunity to share an update after a long time! 🥳 #MIDL2026 just got wrapped up, with another exciting conference in Taiwan! (1/7)

151

Constantin Pape @cppape.bsky.social · 09/07/2026

A big thanks to the AKB and Manfred Eigen Stiftung and everyone else who supported me in Göttingen!

020

Constantin Pape @cppape.bsky.social · 09/07/2026

Sharing a big update: I started a group at the MPINAT in Göttingen! We will develop AI for analyzing how proteins interact in the cellular environment based on cutting edge imaging. This appointment is in parallel to the university, where I will retain my current group.

10587

Constantin Pape @cppape.bsky.social · 05/07/2026

Asked it to do both for distributed graph computation --- way to dangerous ...

000

Constantin Pape @cppape.bsky.social · 05/07/2026

Burned 16% of my Fable usage on prompts that get immediately downgraded to Opus 4.8 😅

140

Reposted by Constantin Pape

Johannes Soltwedel @jo-soltwedel.bsky.social · 03/07/2026

🚨Just out: #NGFF specification 0.6 (rc0): ngff.openmicroscopy.org/specificatio... What's new, what's what? Short thread 🧵👇

ngff.openmicroscopy.org

Version 0.6rc0 — NGFF documentation

2123

Reposted by Constantin Pape

Josh Moore @joshmoore.bsky.social · 03/07/2026

Johannes joined @gerbi-gmb.de with a clear first major task: help lead the completion of #NGFF RFC-5. Today, just before heading off on a well-earned holiday and just shy of his one-year anniversary, 0.6 has a release candidate. Kudos, @jo-soltwedel.bsky.social! 🍻🏝️🚀

074

Constantin Pape @cppape.bsky.social · 30/06/2026

Good to know this happens to others too 😅

010

Constantin Pape @cppape.bsky.social · 30/06/2026

Want to learn about the latest trends and developments in bioimage analysis? Join us for the GLOBIAS workshop in Dortmund this November! I will be hosting a session on foundation models / AI. www.globias.org/activities/g...

globias.org

GloBIAS2026 Workshop Dortmund

GloBIAS 2026 Bioimage Analysts' Workshop Registration is open! Register with this link! 10-13 November, 2026 - Tuesday to Friday Leibniz-Institut für Analytische Wissenschaften – ISAS – e. V. (https...

01510

Constantin Pape @cppape.bsky.social · 23/06/2026

How can we use foundation models such as (micro)SAM to improve electron microscopy segmentation? Check out our new preprint where we found substantial improvements for nucleus, mito, and neurite-segmentation based on initialization and semi-supervised learning with foundation models.

1165

Constantin Pape @cppape.bsky.social · 13/06/2026

Working on library spec stuff with Opus: this model feels so dumb now. I want Fable back! 😢

000

Constantin Pape @cppape.bsky.social · 12/06/2026

Fable 5: amazing model. Safeguards: completely broken and make it unusable for many applications (bio-related and deep learning research, even unrelated to "frontier AI").

150

Reposted by Constantin Pape

Wei-Chung Allen Lee @darbly.bsky.social · 09/06/2026

Now published - the #BANC! A full central nervous system (CNS) connectome of a limbed animal at single-synapse resolution, enabling us to follow sensory-motor arcs and understand how the CNS controls the body. rdcu.be/fncjS. #neuroscience. Video by @quorumetrix.bsky.social 1/18

214984

Reposted by Constantin Pape

CCP-volumeEM @ccp-volumeem.bsky.social · 02/06/2026

Now available on our YouTube channel: Chris Barnes’ CCP-volumeEM Show & Tell webinar, “OME-Zarr: A Next Generation File Format for FAIR Bioimaging Data”. Catch up here: youtu.be/HsqH6RAcJps?...

youtu.be

OME-Zarr: A next generation file format for FAIR bioimaging data with Chris Barnes

YouTube video by CCP-volumeEM

072

Reposted by Constantin Pape

Ali Shaib @alishaib.bsky.social · 01/06/2026

Hello #world, meet 1,000× Expansion Microscopy. A small gel would grow to the size of an Olympic swimming pool, while amino-acid-scale distances become visible with ordinary light microscopy. Led by Helena Hu from @eboyden3.bsky.social's lab, in collab with us. Story: www.biorxiv.org/content/10.6...

1018373

Reposted by Constantin Pape

Virginie Uhlmann @virginieuhlmann.bsky.social · 01/06/2026

📢📢 We are hiring a Bioimage Analyst at the BioVisionCenter! 📢📢 Join our collaboration with the Flow Cytometry Facility and contribute back to the international efforts on open standards for bioimaging data! More information here, reach out if you have any question: jobs.uzh.ch/job-vacancie...

jobs.uzh.ch

UZH: Bioimage Analyst

The University of Zurich is the largest university in Switzerland and one of the most prominent in Europe. For research purposes the Biovision Center and Flow Cytometry Facility offer services to supp...

13428

Reposted by Constantin Pape

Martin Weigert @maweigert.bsky.social · 30/05/2026

Join us again for the Bioimage Computing workshop at ECCV 2026! Paper deadline July 13th 🔥

0135

Reposted by Constantin Pape

Jakob Troidl @jakobtroidl.bsky.social · 27/05/2026

📢 We're recruiting AI Scientists at HHMI Janelia. I'm biased, but these are exceptional opportunities. You'd join as an independent scientist with access to excellent computational infrastructure, some of the best biological data in the world, and a vibrant collaborative environment.

163

Reposted by Constantin Pape

Anna Foix @afoix.bsky.social · 26/05/2026

🔬 One more year! BioImage Computing has been accepted to @eccv.bsky.social 🎉 Show us your work at the intersection of computer vision, ML & biology! Check out our speaker lineup: @arratemunoz.bsky.social, @jakobtroidl.bsky.social & Juliette Griffié 👏 ⏳ Deadline: 13 July ℹ️ www.bioimagecomputing.com

bioimagecomputing.com

BioImage Computing

a truly interdisciplinary workshop

0129

Reposted by Constantin Pape

European Conference on Computer Vision #ECCV2026 @eccv.bsky.social · 20/05/2026

The deadline for submitting #ECCV2026 final reviews is TODAY! Please act responsibly and finalize your discussions and reviews.

043

Reposted by Constantin Pape

Anwai Archit @anwaiarchit.bsky.social · 19/05/2026

At MIDL YRB, we love organizing cool online events that bring young minds in the bio(medical) imaging community together. This time, we’re excited to have @virginieuhlmann.bsky.social share her story! 🥳 Join us tomorrow, 20 May! www.midl.io/yr-storytime @midl-conference.bsky.social

031

Reposted by Constantin Pape

European Conference on Computer Vision #ECCV2026 @eccv.bsky.social · 19/05/2026

We are still a bit more than a day away from the deadline to submit your final #ECCV2026 recommendation. Make sure to discuss your papers with your fellow reviewers before submitting your recommendations!

031

Constantin Pape @cppape.bsky.social · 18/05/2026

Nice to see my name on that list :)

040

Constantin Pape @cppape.bsky.social · 18/05/2026

Sure, no more seeds on the boundaries left ...

150

Constantin Pape @cppape.bsky.social · 18/05/2026

Claude code is so funny. Ask it to generate optimized watershed code: Yes sir! Here's the validated optimized C++ code 10x faster than skimage. Ask it to generate meaningful seeds for the example image: undersegments everything and confidently declares that this are now great parameters.

150

Reposted by Constantin Pape

Simon F. Nørrelykke @simonfn.bsky.social · 06/05/2026

The SwissBIAS Annual Meeting is taking place on June 3rd in Bern. Registration deadline: Friday, May 15th, so make sure to secure your spot now! We look forward to seeing you all there for a fantastic day of science, networking, and inspiration. @swissbias.bsky.social tinyurl.com/SwissBIAS2026

eventbrite.com

SwissBIAS annual meeting 2026

It is the time of the year to make veterans meet new members, discuss to foster feedback on existing activities and promote new initiatives.

095

Constantin Pape @cppape.bsky.social · 03/05/2026

What's the use if I need 5 different API keys from different providers for your tool if you're marginally better than CellPose or microSAM? Either this runs in a self-contained, easy to access (web) environment or I can't see any use coming from this.

060

Constantin Pape @cppape.bsky.social · 03/05/2026

After reviewing a couple of agentic tools for microscopy segmentation: I think this is a bad idea. Setting up these systems is so complicated that it negates any potential productivity gains.

1110

Reposted by Constantin Pape

Peter O'Toole @yorkbioimaging.bsky.social · 24/04/2026

Not just for microscopists, but relevant to all technical specialists. Making the invisible visible: A global examination of careers and recognition for Imaging Scientists in core facilities - Thompson - Journal of Microscopy - Wiley Online Library onlinelibrary.wiley.com/doi/10.1111/...

onlinelibrary.wiley.com

Making the invisible visible: A global examination of careers and recognition for Imaging Scientists in core facilities

Modern life-science research relies heavily on advanced imaging technologies, from microscopes that reveal the structure of cells to imaging systems used in biomedical research. These instruments are...

02113

Reposted by Constantin Pape

Utz Ermel @uermel.bsky.social · 20/04/2026

Excited to share our paper about copick, a dataset API and toolkit for collaborative annotation and analysis of #cryoET data! Whether you're picking particles or curating segmentations, copick reduces friction and brings #OME-Zarr to cryoET without breaking pipelines. 🧵👇 doi.org/10.1002/pro.70578

19230

Reposted by Constantin Pape

Stefan Terjung @stefanterjung.bsky.social · 14/04/2026

We are looking for an experienced light microscopy specialist, motivated to support our users in the Advanced Light Microscopy Facility at EMBL Heidelberg. embl.wd103.myworkdayjobs.com/en-US/EMBL/j... -closing date 11 May 2026 @eurobioimaging.bsky.social @globalbioimaging.bsky.social

embl.wd103.myworkdayjobs.com

03535

Reposted by Constantin Pape

Florian Jug @florianjug.bsky.social · 30/03/2026

🚨Job Alert -- again!🚨 You are into disease risk prediction based on images or other data modalities? This is for you! 🫵 careers.humantechnopole.it/job/Postdoc-... Part of our Scientific AI Flagship @humantechnopole.bsky.social and part of our mission to benefit society via biomedical #AI solutions.

careers.humantechnopole.it

Postdoc on Multimodal AI for Disease Risk Prediction

0128

Reposted by Constantin Pape

Image Cooperative @image.coop · 16/03/2026

Hello, world! We are starting a worker cooperative to build, maintain, teach, and support open source scientific imaging software. image.coop/blog/posts/2...

88444