iMED: A Multi-Endoscope Dataset for Surgical 3D Perception
iMED: A Multi-Endoscope Dataset for Surgical 3D Perception
This paper introduces iMED, the first synchronized multi-endoscope dataset for robot-assisted minimally invasive surgery (RAMIS). The dataset comprises 340 sequences with approximately 170K synchronized timepoints, each containing four views recorded simultaneously from two independent stereo endoscopes across ex vivo, postmortem, and live surgical settings spanning 14 specimens. It includes calibrated camera intrinsics, ArUco-based pose estimates with uncertainty, instrument segmentation masks, and clinical metadata. The authors benchmark 23 state-of-the-art methods using a train-on-one-endoscope, test-on-another protocol, showing that methods relying on photometric supervision degrade under held-out viewpoints while those with geometric regularization remain robust. Large-scale foundation models often outperform domain-specific ones in this setting.
The paper introduces iMED, a dataset of 340 synchronized sequences from two independent stereo endoscopes to benchmark geometric generalization in surgical vision, finding that methods with explicit geometric regularization outperform photometric-only approaches under held-out-viewpoint conditions.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
timepoints
≈170,000
View original evidence
geometric generalization from photometric interpolation along a narrow forward-facing path. iMED provides 340 sequences ~170K synchronized timepoints with 4 views per timepoint) recorded simultaneously from two independent stereo endoscopes across ex vivo, postmortem, and live surgical settings, spanning 14 specimens with diverse anatomical regions and motion regimes. The dataset includes calibrated camera intrinsic
Open source S3sequences
340
View original evidence
geometric generalization from photometric interpolation along a narrow forward-facing path. iMED provides 340 sequences ~170K synchronized timepoints with 4 views per timepoint) recorded simultaneously from two independent stereo endoscopes across ex vivo, postmortem, and live surgical settings, spanning 14 specimens with diverse anatomical regions and motion regimes. The dataset includes calibrated camera intrinsic
Open source S3specimens
14
View original evidence
geometric generalization from photometric interpolation along a narrow forward-facing path. iMED provides 340 sequences ~170K synchronized timepoints with 4 views per timepoint) recorded simultaneously from two independent stereo endoscopes across ex vivo, postmortem, and live surgical settings, spanning 14 specimens with diverse anatomical regions and motion regimes. The dataset includes calibrated camera intrinsic
Open source S3views
4
View original evidence
geometric generalization from photometric interpolation along a narrow forward-facing path. iMED provides 340 sequences ~170K synchronized timepoints with 4 views per timepoint) recorded simultaneously from two independent stereo endoscopes across ex vivo, postmortem, and live surgical settings, spanning 14 specimens with diverse anatomical regions and motion regimes. The dataset includes calibrated camera intrinsic
Open source S3methods
23
View original evidence
s, ArUco-based frame-wise pose estimates with uncertainty quantification, instrument segmentation masks, and rich clinical metadata. Using a train-on-one-endoscope, test-on-another evaluation protocol, we benchmark 23 state-of-the-art methods across rigid and deformable novel view synthesis, pose estimation, feature matching, and monocular depth estimation. Our experiments show that methods relying primarily on photo
Open source S4
Dataset details reported by the source
- Modalities
- instrument segmentation masks Open source S4
- Tasks
- novel view synthesis Open source S4pose estimation Open source S4feature matching Open source S4monocular depth estimation Open source S4
- Embodiments
- two independent stereo endoscopes Open source S3
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
What remains unknown
Dataset licensing and commercial-use rights are not established here.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
geometric generalization from photometric interpolation along a narrow forward-facing path. iMED provides 340 sequences ~170K synchronized timepoints with 4 views per timepoint) recorded simultaneously from two independent stereo endoscopes across ex vivo, postmortem, and live surgical settings, spanning 14 specimens with diverse anatomical regions and motion regimes. The dataset includes calibrated camera intrinsic
Open source S3
s, ArUco-based frame-wise pose estimates with uncertainty quantification, instrument segmentation masks, and rich clinical metadata. Using a train-on-one-endoscope, test-on-another evaluation protocol, we benchmark 23 state-of-the-art methods across rigid and deformable novel view synthesis, pose estimation, feature matching, and monocular depth estimation. Our experiments show that methods relying primarily on photo
Open source S4
iMED: A Multi-Endoscope Dataset for Surgical 3D Perception
Open source S1
Implications for data suppliers
RoboSignal interpretation and collection questions, not statements of buyer demand.
- Confirm the required data type and collection setting with the buyer; this source does not establish a complete collection specification.
- Compare the reported units and scope before using these quantities in a budget. Recording hours, sensor-hours and trajectories are different measures.
- Check the exact dataset version, license, modalities and task coverage against the intended use.
- Validate demand and acceptance criteria with a buyer before scaling. Publication, popularity and a research result do not establish a purchase commitment.
Source：arXiv Robotics — research abstracts · arxiv.org