Studying 3D vision problems for robotic surgery. PhD Candidate @TumCamp | Senior Research Engineer @ImFusionGmbH 🟦
Munich, Bavaria
Joined May 2011
- Tweets116
- Following893
- Followers415
- Likes10.4K
Pinned Tweet
Working on a new point tracking method or need a faster alternative?
Check out LiteTracker — now with results on natural, non-surgical, benchmarks posted on the repository!
It remains highly competitive at a fraction of the latency.
👉 github.com/ImFusionGmbH/lite…
Mert Karaoglu retweeted
Future Dynamic 3D Reconstruction: A 3D World Model with Disentangled Ego-Motion
Nils Morbitzer, Jonathan Evers, Artem Savkin, Thomas Stauner, @NassirNavab, @fedassa, @ste_gasperini
tl;dr: decouple ego-motion from world-motion; latent proxy->ego-motion
arxiv.org/abs/2606.18250
Mert Karaoglu retweeted
Hybrid Foveated Path Tracing with Peripheral Gaussians for Immersive Anatomy
Contributions:
• A novel hybrid rendering pipeline for anatomy integrating streamed foveated path tracing and a regenerating 3DGS peripheral cloud model
• A depth-guided reprojection method for the path-traced anatomy to hide latency and decouple render speed from display refresh
• An evaluation exploring interactive peripheral model generation, system performance, and perceptual fidelity compared to standalone path tracing and 3DGS
Mert Karaoglu retweeted
STIR Challenge 2025 is a wrap!
Slides here: synapse.org/Synapse:syn65877…
If you submitted and want to update, or if you didn't make it in time: you have until December 15th to participate!
Thanks to our sponsors @IntuitiveSurg @ImFusionGmbH , and my co-organizer, @mertkaraoglu
STIR Challenge 2025 update! We have extended the deadline to August 31.
Submission details at stir-challenge.github.io/
🧵LiteTracker is accepted to MICCAI 2025!
LiteTracker is a low-latency tissue tracking method designed for real-time surgical applications.
Key highlights:
🎞️ Frame-by-frame, low-latency tracking on commercial GPUs
🚀 >7× faster than its predecessor and >2× faster than current SOTA
🤝 Compatible with CoTracker3 weights — no retraining needed
📊 High accuracy tracking & occlusion prediction on STIR and SuPer
📄 arXiv: arxiv.org/abs/2504.09904
💻 Code: github.com/ImFusionGmbH/lite…
Many thanks to all collaborators at @ImFusionGmbH and @TumCamp : Wenbo Ji, Ahmed Abbas, @NassirNavab , @BusamBenjamin , and Alexander Ladikos.
See you all in Daejeon 🇰🇷
#MICCAI2025
🧵STIR Challenge returns for MICCAI 2025
We're inviting researchers to submit their point tracking methods on our challenging surgical tissue tracking benchmark.
If you're working on robust and efficient point tracking, this is a great opportunity to benchmark and compete.
🔗 Details on the submission, tasks and prizes: stir-challenge.github.io
🔗 Dataset from last year's challenge: stir-challenge.github.io//st…
#MICCAI2025 #EndoVis #STIRChallenge
🔗 STIR Dataset: arxiv.org/abs/2309.16782
🔗 Baseline results on the last year's challenge:
- arxiv.org/abs/2503.24306
- arxiv.org/abs/2504.09904
Many thanks to our sponsors @ImFusionGmbH and @IntuitiveSurg for their support and to @MICCAI_Society and @endo_vis for their hosting!
Mert Karaoglu retweeted
2025 STIR challenge update:
- prizes posted
- validation is open
Modify, train, tweak any tracker of your choice and submit it!
Register at synapse.org/Synapse:syn65877… to participate
Final submissions are due August 20.
#MICCAI2025 #EndoVis @endo_vis
@IntuitiveSurg
@ImFusionGmbH
📢 Exciting news! Check out DynaMoN, our new paper on motion-aware, robust camera localization for dynamic neural radiance fields, now accepted to
#IEEE RA-L!
Huge thanks to the team for the amazing effort!
👉 hannahhaensen.github.io/Dyna…
@ImFusionGmbH @TumCamp @TUM_MIRMI #FAU
Mert Karaoglu retweeted
🚨🚨🚨🚨🚨 Paper Alert 🚨🚨🚨🚨🚨
NeRFtrinsic Four: An End-To-End Trainable NeRF Jointly Optimizing Diverse Intrinsic and Extrinsic Camera Parameters has been accepted to #CVIU!!!
(1/4)
📢 The challenge is officially over, but we are still open for submissions! Send in your method until Dec. 15 to contribute to our benchmark. Post-MICCAI submissions will also be invited to join in the summary paper.
🔗 For more details: stir-challenge.github.io//pa…
📢 Exciting news! We are organizing the STIR point tracking challenge this year at #MICCAI2024 @MICCAI_Society as a part of @endo_vis.
🔗stir-challenge.github.io/
Mert Karaoglu retweeted
🎉 The STIR Challenge presented this Oct. 10th in #endovis at #MICCAI2024
A huge thanks to co-organizers @mertkaraoglu, Soham & Jonathan; sponsors @ImFusionGmbH @nvidia @IntuitiveSurg, and participants!
📊 Results + slides are on our synapse page: synapse.org/Synapse:syn54126…
It was a pleasure to organize the STIR challenge at @MICCAI_Society 2024 with @adam_thaddius and Jonathan McLeod. Many thanks to our sponsors @ImFusionGmbH, @IntuitiveSurg, and @nvidia, as well as @endo_vis for their great support and hosting.
Congratulations to the winners!
Mert Karaoglu retweeted
📣 We'll be presenting our point tracking challenge, STIR (stir-challenge.github.io/) tomorrow, October 10th, at #MICCAI24 as part of the #endovis challenge! We've attached a map.
Opening remarks: 8:00am
STIR: 8:10am
Sponsors:
@IntuitiveSurg, @ImFusionGmbH, @nvidia
Mert Karaoglu retweeted
Point tracking in surgery? Get ready for the STIR Challenge! We're presenting this Thursday, October 10. as part of @endo_vis at #MICCAI24
EndoVis opening remarks are at 8am, and we start at 8:10am.
stir-challenge.github.io/
