Sophia Bano
UCL, United Kingdom
Robot vision and scene understanding for minimally invasive surgery.
Data Curation
& Augmentation
in Medical Imaging.
An ECCV workshop on data as the bottleneck for robust surgical computer vision - from curated multi-endoscope datasets to large multimodal clinical benchmarks and surgical visual understanding.
IIEditionAfter CVPR 2024
3DatasetsiMED | CLIMB | SurgVU
4Invited speakersConfirmed program
8-9SeptAt ECCV 2026
§ 01About the workshop
PP. 01 - 04
From scarce data
to safer medicine.
A focused program of invited talks, peer-reviewed papers, and dataset spotlights.
Reliable medical systems depend on data quality as much as model novelty. Intraoperative video is scarce, expensive to annotate, and ethically restricted. DCA-MI advances data-centric methods and benchmarks as first-class research contributions, with particular attention to surgical scene understanding, endoscopic geometry, and clinical translation.
"The bottleneck is rarely the model. It is almost always the data."
Organizing Committee
§ 02Featured speakers
Invited speakers
Invited speakers
for DCA-MI 2026.
Lena Maier-Hein
DKFZ / Heidelberg University, Germany
Surgical data science, benchmarking, and reproducible evaluation.
José M. M. Montiel
Universidad de Zaragoza, Spain
Visual SLAM, deformable SLAM for endoscopy, EndoMapper.
Mengya Xu
CUHK, Hong Kong
Medical and surgical AI across MICCAI, IPCAI, and ICRA.§ 03Areas of focus
Aligned with invited talks
What we want
to talk about.
Robot vision and surgical scene understanding
How robust visual perception supports minimally invasive surgical workflows.
Surgical data science and benchmarking
How dataset design, provenance, and reproducible evaluation shape trustworthy medical AI.
Endoscopic geometry, SLAM, and neural rendering
How calibrated surgical data enables mapping, reconstruction, and deformable scene representations.
Autonomous surgery and clinical translation
How surgical datasets connect perception research to deployable clinical systems.
§ 04Advisory board
Confirmed advisors and invited speakers
Featured advisor
Stefanie Speidel
NCT Dresden professor working on surgical data science, computer-assisted surgery, robotic vision, and AI-enabled clinical translation.
ProfileInvited speakerSophia Bano
UCL researcher focused on robot vision and scene understanding for minimally invasive surgery.
Invited speakerLena Maier-Hein
DKFZ and Heidelberg University researcher in surgical data science, benchmarking, and reproducible evaluation.
Invited speakerJosé M. M. Montiel
Universidad de Zaragoza researcher known for visual SLAM, deformable SLAM for endoscopy, and EndoMapper.
Invited speakerMengya Xu
Chinese University of Hong Kong researcher working on medical and surgical AI.
§ 05Featured datasets
Curation case studies
Three benchmarks,
three data lessons.
Each spotlight treats a dataset as a research contribution: acquisition decisions, ground-truth design, split strategy, and evaluation pitfalls.
PRIMARY SPOTLIGHT
iMED
A Multi-Endoscope Dataset for Surgical 3D Perception
340 sequences~170K frame pairs60 FPS19 methods10 ex vivo organs18 cadaveric scenes10 live porcine procedures4 camera streams
iMED focuses the workshop on synchronized, dual-view, deformable, specular surgical scenes. The dataset is designed to stress test visual geometry, photogrammetry, and endoscopic reconstruction beyond rigid-world assumptions.
FEATURED DATASET
CLIMB
Clinical Large-scale Integrative Multimodal Benchmark
4.51M samples19.01TB33 institutions13 domains
A multimodal clinical benchmark spanning imaging, language, time series, graph, and multimodal patient data.
arXiv:2503.07667
FEATURED DATASET
SurgVU
Surgical Visual Understanding Dataset
840 hours~18M frames280 clips12 tools155 sessions8 tasks
Robot-assisted surgical video with tool and task labels, built for visual understanding under operating-room scale and imbalance.
arXiv:2501.09209Figure sources: iMED local paper assets; CLIMB Figure 2 from arXiv:2503.07667; SurgVU Figure 1 from arXiv:2501.09209.
§ 06Calendar
All times AOE
Important
dates.
15 May 2026Call for papers postedOpen
01 Jul 2026Paper submission deadlineHard
01 Aug 2026Notification of acceptanceEmail
15 Aug 2026Camera-ready dueFinal
08-09 Sep 2026Workshop at ECCV 2026In person
§ 02Featured DatasetiMED | CLIMB | SurgVU
Dataset
spotlights.
Three benchmarks,
three data lessons.
Each spotlight treats a dataset as a research contribution: acquisition decisions, ground-truth design, split strategy, and evaluation pitfalls.
PRIMARY SPOTLIGHT
iMED
A Multi-Endoscope Dataset for Surgical 3D Perception
340 sequences~170K frame pairs60 FPS19 methods10 ex vivo organs18 cadaveric scenes10 live porcine procedures4 camera streams
iMED focuses the workshop on synchronized, dual-view, deformable, specular surgical scenes. The dataset is designed to stress test visual geometry, photogrammetry, and endoscopic reconstruction beyond rigid-world assumptions.
FEATURED DATASET
CLIMB
Clinical Large-scale Integrative Multimodal Benchmark
4.51M samples19.01TB33 institutions13 domains
A multimodal clinical benchmark spanning imaging, language, time series, graph, and multimodal patient data.
arXiv:2503.07667
FEATURED DATASET
SurgVU
Surgical Visual Understanding Dataset
840 hours~18M frames280 clips12 tools155 sessions8 tasks
Robot-assisted surgical video with tool and task labels, built for visual understanding under operating-room scale and imbalance.
arXiv:2501.09209Figure sources: iMED local paper assets; CLIMB Figure 2 from arXiv:2503.07667; SurgVU Figure 1 from arXiv:2501.09209.
§ 02Important DatesAll times AOE
Calendar
at a glance.
Live countdownCalculating...
Call for papers posted
OpenReview information, templates, and final submission instructions become available.
Paper deadline
Full papers, extended abstracts, and dataset submissions due by 23:59 AOE.
Acceptance notifications
Decisions sent by email with presentation type.
Camera-ready due
Final PDFs uploaded and program frozen.
Workshop at ECCV
DCA-MI 2026 meets during ECCV on September 8-9.
§ 03Call for PapersOpen Spring 2026
Submit your
research.
The data itself is the subject.
We invite contributions across curation, augmentation, restoration, 3D perception, learning with limited or imperfect data, and clinical translation. Submissions may be empirical, methodological, position-style, or new datasets and benchmarks with reproducible baselines.
Cross-theme work is welcome.
We are especially interested in work that shows how upstream data decisions propagate into downstream clinical and surgical performance.
Three submission tracks.
Full papers, extended abstracts, and dataset / benchmark submissions are welcome. Final page limits and template links will be posted with the OpenReview call.
Double-blind review.
Each submission receives technical and domain review, with dataset papers evaluated for provenance, license clarity, and reproducibility.
Provenance is a first-class concern.
All datasets must document source, consent basis, licensing, and known coverage limits. Submissions with unclear protected-data handling may be desk-rejected.
§ 05ProgramSchedule draft
Workshop
schedule.
05ASchedule
Time | Session | Format
TimeEventType
Opening remarks
Welcome and workshop overview from the organizers.
Keynote 1: Dataset scarcity, design and curation in medical imaging
Invited talk on the data bottleneck for robust medical and surgical AI.
Coffee Break & Poster Session I
Interactive poster session and attendee discussion.
Oral Session 1: 3D Understanding and Spatial Data
Four accepted papers, 15 minutes each.
Break
Lunch break.
Keynote 2: SLAM, Neural Rendering
Invited talk on endoscopic geometry, mapping, and scene representation.
Data Curation for Challenge Session
iMED: Multi-Endoscope Dataset | 20 min
CLiMB: Benchmark for Colonoscopy SLAM | 20 min
SurgVU: Surgical Visual Understanding | 20 min
Coffee Break & Poster Session II
Second poster block and hallway discussion.
Keynote 3: Autonomous Surgery and Clinical Translation
Invited talk on translating surgical data into robust clinical systems.
Oral Session 2: Visual degradation and restoration
Four accepted papers, 15 minutes each.
Closing Remarks
Closing notes and next steps from the organizers.
05BSpeakers
Invited speakers
Sophia Bano
UCL, United Kingdom
Robot vision and scene understanding for minimally invasive surgery.Lena Maier-Hein
DKFZ / Heidelberg University, Germany
Dataset scarcity, design, curation, and reproducible surgical data science.José M. M. Montiel
Universidad de Zaragoza, Spain
SLAM, neural rendering, deformable reconstruction, and EndoMapper.Mengya Xu
CUHK, Hong Kong
Autonomous surgery, clinical translation, and medical computer vision.§ 06Organizers8 organizers
The team
behind it.
Fengyi Jiang
Primary contact | Intuitive Surgical
Sierra Bonilla
iMED dataset lead | UCL Hawkes Institute
Javier Morlana
CLiMB benchmark lead | Universidad de Zaragoza
Ray Zhang
Organizer | Intuitive Surgical
Mary Jin
Organizer | Intuitive Surgical
Shuoqi Chen
Computer Vision & Medical Imaging engineer at Intuitive Surgical, specializing in advanced imaging and robotic-assisted procedures. CMU Robotics Institute alumnus; presenter and reviewer across IEEE TRO, IROS, CVPR, and ICML.
shuoqi.chen@intusurg.com
Jingpei Lu
Research Scientist | Intuitive Surgical
Rogerio Nespolo
SurgVU lead | Intuitive Surgical
07AAdvisory Board
Confirmed advisors and invited speakers
Featured advisor
Stefanie Speidel
Professor for Translational Surgical Oncology at NCT Dresden, working on surgical data science, computer-assisted surgery, robotic vision, and AI-enabled clinical translation.
Profile
Robot vision
Sophia Bano
UCL researcher focused on robot vision and scene understanding for minimally invasive surgery.
Profile
Surgical data science
Lena Maier-Hein
DKFZ and Heidelberg University researcher in surgical data science, benchmarking, and reproducible evaluation in medical AI.
Profile
Visual SLAM
José M. M. Montiel
Universidad de Zaragoza researcher known for visual SLAM, ORB-SLAM, deformable SLAM for endoscopy, and the EndoMapper dataset.
Profile
Medical AI
Mengya Xu
Chinese University of Hong Kong researcher working on medical and surgical AI, with recent work across MICCAI, IPCAI, and ICRA.
Profile