Excited to present two works at #CVPR2025! If you're attending CVPR, feel free to stop by and say hi! 📍 Main Conference MM-OR: A Large Multimodal OR Dataset for Semantic Surgical Scene Understanding Arxiv: https://t.co/85eZUlksoS Poster Session: 4 Poster ID: 341

This work was co-led with @arda_mamur , and realized together with our fantastic co-authors Felix Tristram, @ChPellegrini , @aiiakisi , @BusamBenjamin, and @NassirNavab . A huge thanks to them and to everyone who supported us along the way.

We believe EgoExOR will catalyze innovation in OR perception, paving the way for smarter, safer surgical environments, with implications for any field requiring precise human-robot collaboration in complex environments.

Captured with up to five egocentric Meta ARIA glasses and five external RGB-D cameras, ultrasound monitor feed, audio, fused 3D point-cloud reconstructions, gaze, hand-pose signals, and detailed scene graph annotations, it provides a comprehensive view of surgical workflows.

EgoExOR is the first operating room dataset that fuses egocentric and exocentric perspectives, covering two emulated spine procedures for spinal pain relief and herniated disc removal: Ultrasound-Guided Needle Insertion and Minimally Invasive Spine Surgery.

🎉 Excited to share that our paper “EgoExOR: An Ego–Exo-Centric Operating Room Dataset for Surgical Activity Understanding” has been accepted to #NeurIPS2025. #SDS #Egocentric #SceneGraphs #OperatingRoom Paper: https://t.co/YOKuuv3Edg Code & Data: https://t.co/wIj0ml7INe

📍 Multimodal Learning & Applications Workshop LF-SGG: Location-Free Scene Graph Generation 🔗

I am happy to share that RaDialog was accepted to MIDL 2025! We look forward to presenting the newest version of RaDialog, with an improved architecture, extended instruct dataset, more thorough evaluation and analysis and Hugging Face integration.

🎉 Thrilled to share MM-OR has been accepted to #CVPR2025. MM-OR is a new, large scale multimodal operating room dataset for semantic understanding of surgeries. It captures robotic knee replacement surgeries with multi-view RGB-D video, audio, speech transcripts, robotic logs,

We believe that our new task and its objective evaluation paves the way to more efficient scene interpretation from visual content in the form of scene graphs.

We demonstrate the effectiveness of our method on three scene graph generation datasets as well as two downstream tasks, image retrieval and visual question answering, and show that our approach is competitive to existing methods while not relying on location cues.

We design the first LF-SGG method, Pix2SG, using autoregressive sequence modeling and to objectively evaluate LF-SGG, we design and implement an efficient branching algorithm.

Recognizing that many downstream application do not require location data (image retrieval, visual question answering, image captioning, and more), we break this dependency and introduce location-free scene graph generation (LF-SGG).

Traditional scene graph generation methods rely on location labels in form of bounding boxes or segmentation masks, increasing annotation costs and limiting dataset expansion.

Excited to introduce the task of Location-Free Scene Graph Generation (LF-SGG). To this end, we design the first LF-SGG method, Pix2SG and the corresponding evaluation method based on heuristic tree search. Paper: https://t.co/S92na8xCUP Code: https://t.co/8I4g9ounjG

Dear all, This is the last magazine by RSIP Vision: https://t.co/ANqR0V5JoL The company has sent 202 magazines since April 2016. Computer Vision News will continue its publication. Enjoy the reading and... don't miss the cat on page 43! Ralph

Title: ORacle: Large Vision-Language Models for Knowledge-Guided Holistic OR Domain Modeling Paper Link: https://t.co/NCO7wXOClN Authors: Ege Özsoy*, Chantal Pellegrini*, Matthias Keicher, Nassir Navab Code:  https://t.co/JvE505knDX

We are extremely honered that our work“ORacle“ has received the Best Paper Runner-Up Award at MICCAI 2024! Huge thanks to our supervisors @MatthiasKeicher @NassirNavab #MICCAI2024 #BestPaperRunnerUp

Another very well-deserved oral from E. Özsoy from @NassirNavab lab. Really enjoying the current #miccai2024 session. https://t.co/wZMtxYissk

Poster Number: W-PM-188 Poster Session 6, Computer Assisted Interventions and Surgery @MICCAI_Society @MiccaiStudents