🌺 Just 4 days to go!
Join us in Honolulu for the Instance-Level Recognition and Generation Workshop at #ICCV2025 🏝
🗓️ Oct 19, 8:30am–12:30pm 📍 Room 306 A

We’ll have amazing keynotes, plus oral and poster sessions featuring accepted and invited papers.
Don’t miss it!
ilr-workshop.github.io/ICCVW2025/

• SigLIP2 → still the best for text-to-image retrieval.
– Unlike SigLIP models, PE shows a large gap between its image-to-image and text-to-image performance.

• PE → achieves state-of-the-art in image-to-image retrieval with linear adaptation.
  – Gains +12% performance from adaptation.
  – Outperforms SigLIP2 (previous best) by +2%.
  – DINOv3 lags behind, gaining less than +2% from adaptation

🔍 Key results from our new evaluation on ILIAS:

• DINOv3 → sets a new state-of-the-art in image-to-image retrieval without linear adaptation
  – The large variant outperforms all other models by a significant margin.
  – The base outperforms the large variants of other model series.

🚀 new state-of-the-art on ILIAS dataset!

Curious how well the latest models can recognize particular objects?
We evaluated the base and large variants of DINOv3 and Perception Encoder (PE) on instance-level image retrieval.

See the results 👉 vrg.fel.cvut.cz/ilias/

🚨 Deadline Extension

Instance-Level Recognition and Generation (ILR+G) Workshop at ICCV2025 @iccv.bsky.social

📅 new deadline: June 26, 2025 (23:59 AoE)
📄 paper submission: cmt3.research.microsoft.com/ILRnG2025
🌐 ILR+G website: ilr-workshop.github.io/ICCVW2025/

#ICCV2025 #ComputerVision #AI

Are you at @cvprconference.bsky.social? Come by our poster!
📅 Sat 14/6, 10:30-12:30
📍 Poster #395, ExHall D

Local representation is a key ingredient, while global representation, despite being efficient and compact, performs much lower

ILIAS is very challenging and multi-domain, where single domain instance-level retrieval models fail to perform robustly

SigLIP achieves the best performance. VLMs are the top-performing models in general. Also, the performance gap between image-to-image and text-to-image models is surprisingly small!

It consists of 1,000 object instance depicted in 5,947 manually collected images - 1,232 query images and 4,715 positives - 1,000 text queries. It is combined with 100M distractor from YFCC100M. It includes positives in very challenging conditions and distractor with similar appearance or semantics

ILIAS is a large-scale test dataset for evaluation on Instance-Level Image retrieval At Scale. It is designed to support future research in image-to-image and text-to-image retrieval for particular objects and serves as a benchmark for evaluating foundation models and retrieval techniques.