Omiros Pantazis

AI PhD @UCL. Research focus on enhancing scientific discovery in the intersection of #computer_vision & #biodiversity_monitoring. Part of @Biome_Health project.

[Georgios Tzannetos @gtzannet]

Excited our work got accepted at #NeurIPS2025 last week 🎉

We design curriculum strategies to train models under test-time constraints.
👉 Result: LLMs that solve math reasoning with compressed CoT — same accuracy, faster inference 🚀

More details soon
See you in San Diego ✈️

[FGVC Workshop @fgvcworkshop]

FGVC12 Workshop accepted to CVPR 2025, Nashville!
CALL FOR PAPERS: https://t.co/60U5VTaYEl
We discuss domains where expert knowledge is typically required and investigate artificial systems that can distinguish numerous very similar visual concepts.
#CVPR #CVPR2025 #AI

[Omiros Pantazis @panom1]

Humbled to see @MIT spotlighting our research on language-based image retrieval!

Through this work we created a retrieval benchmark to test AI models under a challenging biodiversity monitoring setting, aiming to inspire new methods and accelerate scientific discovery.

[MIT CSAIL @MIT_CSAIL]

Could multimodal vision language models (VLMs) help biodiversity researchers retrieve images for their studies? 🤔

MIT CSAIL, @ucl, @inaturalist, @EdinburghUni, & @UMassAmherst researchers designed a performance test to find out. Each VLM’s task: Locate & reorganize the most relevant results w/i the team’s "INQUIRE" dataset, composed of five million wildlife pictures & 250 search prompts from ecologists & other biodiversity experts: https://t.co/clOccGg7G3

[Eddie Vendrow @EdwardVendrow]

I'm excited to present INQUIRE at #NeurIPS2024 this week!

DM me or drop by to catch up about anything-- AI for scientific discovery, synthetic data, reasoning, reliability, etc!

🕓 Friday 11am-2pm
📷East Exhibit Hall A-C #4510

[Oisin Mac Aodha @oisinmacaodha]

Interested in species distribution modelling? If so, check out our recent work where we introduce a new model that can estimate the spatial range of a species from only a text description as input.

Max Hamilton will be at #NeurIPS2024 this week to present it.

[Sara Beery @sarameghanbeery]

I'm incredibly honored to be a part of this cohort of amazing researchers working on hard problems in AI 💜

[Sara Beery @sarameghanbeery]

Amazing work led by my student @EdwardVendrow and @panom1 from @ucl on enabling scientists to interactively probe massive-scale data on @inaturalist via text

If you'll be at #NeurIPS2024 next week come learn more at our poster!

Dec. 13th, 11am-2pm, East Exhibit Hall A-C #4510

[Oisin Mac Aodha @oisinmacaodha]

Check out Eddie and Omiros' work introducing our new vision-language retrieval dataset INQUIRE which will be presented at #NeurIPS2024 next week.
@EdwardVendrow @panom1

[Omiros Pantazis @panom1]

Check out INQUIRE, our new text-to-image retrieval dataset that challenges multimodal AI models. The paper introducing the dataset and benchmarks will be presented at #NeurIPS2024 next week!

[Eddie Vendrow @EdwardVendrow]

🎯 How can we empower scientific discovery in millions of nature photos?

Introducing INQUIRE: A benchmark testing if AI vision-language models can help scientists find biodiversity patterns- from disease symptoms to rare behaviors- hidden in vast image collections.

Thread👇🧵

[Oisin Mac Aodha @oisinmacaodha]

In Generalized Category Discovery (GCD), the goal is to train models that can discover new categories using knowledge learned from a labelled dataset.

[Eddie Vendrow @EdwardVendrow]

🔬Scientific discovery is one of the most exciting use cases for AI.

Today at the ECCV @EvalFomo workshop I’ll present a poster on our newly accepted @NeurIPSConf paper on benchmarking scientific discovery in image collections. Come by and stay tuned for more!

📍Amber 5, 3:15

[Omiros Pantazis @panom1]

Are you in Milan attending @eccvconf and interested in learning what #MultimodalAI can do for natural world image retrieval? Join the "Emergent Visual Abilities and Limits of Foundation Models" workshop where @EdwardVendrow will be presenting our @NeurIPSConf accepted paper

[Sara Beery @sarameghanbeery]

@EdwardVendrow will be presenting a poster at the evaluating foundation models workshop on our newly accepted @NeurIPSConf paper on benchmarking scientific discovery in image collections

[devistuia @devistuia]

If you are in Milan at @eccvconf and what to chat about it, please reach out! We are still welcoming applications!

[Sara Beery @sarameghanbeery]

Interested in a PhD or Masters at the intersection of AI and Ecology? This research area is growing rapidly, and it can be hard to figure out which research groups where are doing what! Come join our infosession to hear from PIs worldwide about their research and goals!!

[Yulia Gryaditskaya @ygryaditskaya]

🎉 At ECCV'24 Gizem Unlu @gizemesraunlu will present our work "GroundUp" introducing a sketch-based ideation tool for 3D city massing. We step towards empowering architects to easily switch between 2D sketches and 3D models, making iteration and idea sharing smoother!

Details ⬇️

[Oisin Mac Aodha @oisinmacaodha]

We have been working with the @iNaturalist team to investigate vision language models for fine-grained image retrieval. They now have a cool demo online.

You can read more about it in this blog post:
https://t.co/hpDyGiosKI

[Sara Beery @sarameghanbeery]

A first demo of this line of work has just been launched on @iNaturalist 🎉🎉🎉

You can read more about it here https://t.co/dMaqv5nkER and on their blog: https://t.co/QdDHEmjzL9

[Sara Beery @sarameghanbeery]

Happy to be at #CVPR2024 in my hometown! I'll be giving a talk today at the CV4Science workshop at 4:15 on interactive query systems for scientific discovery, then moderating a panel at the #CV4Animals workshop at 5! Then I'll be hosting an #AIforConservation meetup at 6:30 💜💜

[Kate E. Jones 🦇 @ProfKateJones]

!!!JOB ALERT!!! Assistant/Associate Prof in One Health @ucl. We are looking for a fabulous new colleague @UCLCBER for our People & Nature Lab @UCLEast - deadline 10th July https://t.co/lNAxjYytwe