Here a cool student who became a cool TA and master of Infrastructure 😃

We’ll be presenting at the #BIC workshop at @iccv.bsky.social! This year’s BIC has amazing speakers 🌺 — join us on 19 October at the Honolulu Convention Center to explore cutting-edge methods at the intersection of computer vision and biology 🔬 #ICCV2025 More details: www.bioimagecomputing.com

Thanks to @alex-krull.bsky.social and @unibirmingham.bsky.social for supporting the interdisciplinary workshop between computer vision and biology, and for helping us secure a space in computer science venues 🔬👩‍💻 Looking forward to hearing all the contributors and seeing their posters! #ICCV25

We’ll be presenting at the #BIC workshop at @iccv.bsky.social! This year’s BIC has amazing speakers 🌺 — join us on 19 October at the Honolulu Convention Center to explore cutting-edge methods at the intersection of computer vision and biology 🔬 #ICCV2025 More details: www.bioimagecomputing.com

Also don’t miss:
🧬👯 @epetsalaki.bsky.social’s talk “From Data to Digital Twins” (Wed, Digital Twin Models)

📊🧬 My poster P03.14 – Towards Data-Driven Executable Models (Wed)

🧩🤖 @alussana.bsky.social’s poster P18.13 – Data-Driven Inference of Functionally Coherent Gene Modules (Mon)

Thank you @zaminiqbal.bsky.social and for all your orevious encouraging! I am sure is going amazing in Bath!

@alex-krull.bsky.social thanks for be such an inspiration and huge support during our journey, more to come together 😄!

I couldn't have asked for a better supervisor, inspiration and support. ShapeEmbed is the proof that science is better when you have team work 👩‍💻👩‍💻 thanks @virginieuhlmann.bsky.social

A huge thank you 💙 to my supervisor @virginieuhlmann.bsky.social, my co-advisor @alex-krull.bsky.social and my colleague Craig Russell who shared feedback and ideas. And to the community already using ShapeEmbed — it’s inspiring to see it applied in different biological contexts! (N/N)

Why is ShapeEmbed useful in your research? It can extract shape descriptors to distinguish cell phenotypes and reveal drug-condition induced changes through clustering and latent-space exploration, offering a powerful tool for biological shape analysis. (5/N)

On top of all of this, the latent space learned by ShapeEmbed is generative: it can be used to generate shapes and compute meaningful shape averages and trajectories, enabling data exploration & visualization—which are useful for both discovery and hypothesis testing. (4/N)

ShapeEmbed is performant even with very limited data — a huge advantage in biological context where images are scarce and may be costly to acquire. Small microscopy sets are enough to learn useful shape descriptors. (3/N)

ShapeEmbed encodes Euclidean distance matrices relying on a novel encoder design and a new indexation-invariant loss. Through this, it learns highly-informative shape features that surpass classical approaches and current SotA learning methods. (2/N)