On the Role of Depth in Surgical Vision Foundation Models: An Empirical Study of RGB-D Pre-training

• URL: http://arxiv.org/abs/2601.18929v1
• Date: Mon, 26 Jan 2026 20:04:57 GMT
• Title: On the Role of Depth in Surgical Vision Foundation Models: An Empirical Study of RGB-D Pre-training
• Authors: John J. Han, Adam Schmidt, Muhammad Abdullah Jamal, Chinedu Nwoye, Anita Rau, Jie Ying Wu, Omid Mohareri,
• Abstract summary: Vision foundation models (VFMs) have emerged as powerful tools for surgical scene understanding.<n>Current approaches predominantly rely on unimodal RGB pre-training, overlooking the complex 3D geometry inherent to surgical environments.<n>We conduct a large-scale empirical study comparing eight ViT-based VFMs that differ in pre-training domain, learning objective, and input modality.
• Score: 7.4397612353762606
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Vision foundation models (VFMs) have emerged as powerful tools for surgical scene understanding. However, current approaches predominantly rely on unimodal RGB pre-training, overlooking the complex 3D geometry inherent to surgical environments. Although several architectures support multimodal or geometry-aware inputs in general computer vision, the benefits of incorporating depth information in surgical settings remain underexplored. We conduct a large-scale empirical study comparing eight ViT-based VFMs that differ in pre-training domain, learning objective, and input modality (RGB vs. RGB-D). For pre-training, we use a curated dataset of 1.4 million robotic surgical images paired with depth maps generated from an off-the-shelf network. We evaluate these models under both frozen-backbone and end-to-end fine-tuning protocols across eight surgical datasets spanning object detection, segmentation, depth estimation, and pose estimation. Our experiments yield several consistent findings. Models incorporating explicit geometric tokenization, such as MultiMAE, substantially outperform unimodal baselines across all tasks. Notably, geometric-aware pre-training enables remarkable data efficiency: models fine-tuned on just 25% of labeled data consistently surpass RGB-only models trained on the full dataset. Importantly, these gains require no architectural or runtime changes at inference; depth is used only during pre-training, making adoption straightforward. These findings suggest that multimodal pre-training offers a viable path towards building more capable surgical vision systems.

Related papers

• DeFM: Learning Foundation Representations from Depth for Robotics [49.77188649197404]
  We present DeFM, a self-supervised foundation model trained entirely on depth images for robotic applications.<n>DeFM learns geometric and semantic representations that generalize to diverse environments, tasks, and sensors.<n>It achieves state-of-the-art performance and demonstrates strong generalization from simulation to real-world environments.
  arXiv  Detail & Related papers  (2026-01-26T19:45:31Z)
• TAP-CT: 3D Task-Agnostic Pretraining of Computed Tomography Foundation Models [39.00742360251856]
  Existing foundation models (FMs) in the medical domain often require extensive fine-tuning or rely on training resource-intensive decoders.<n>We introduce a suite of task-agnostic pretraining of CT foundation models (TAP-CT)<n>Our approach incorporates targeted modifications to patch embeddings, positional encodings, and volumetric augmentations, making the architecture depth-aware.
  arXiv  Detail & Related papers  (2025-11-30T12:43:15Z)
• Learning to Efficiently Adapt Foundation Models for Self-Supervised Endoscopic 3D Scene Reconstruction from Any Cameras [41.985581990753765]
  We introduce Endo3DAC, a unified framework for endoscopic scene reconstruction.<n>We design an integrated network capable of simultaneously estimating depth maps, relative poses, and camera intrinsic parameters.<n>Experiments across four endoscopic datasets demonstrate that Endo3DAC significantly outperforms other state-of-the-art methods.
  arXiv  Detail & Related papers  (2025-03-20T07:49:04Z)
• SegBook: A Simple Baseline and Cookbook for Volumetric Medical Image Segmentation [20.026663367994356]
  Large amounts of full-body CT images provide the opportunity to pre-train powerful models. It remains unclear in which conditions these pre-trained models can be transferred to various downstream medical segmentation tasks. We collected 87 public datasets varying in modality, target, and sample size to evaluate the transfer ability of full-body CT pre-trained models.
  arXiv  Detail & Related papers  (2024-11-21T19:00:01Z)
• Evaluating Pre-trained Convolutional Neural Networks and Foundation Models as Feature Extractors for Content-based Medical Image Retrieval [0.37478492878307323]
  Content-based medical image retrieval (CBMIR) depends on image features, which can be extracted automatically or semi-automatically.<n>In this study, we used several pre-trained feature extractors from well-known pre-trained convolutional neural networks (CNNs) and pre-trained foundation models.<n>Our results show that, overall, for the 2D datasets, foundation models deliver superior performance by a large margin compared to CNNs.<n>Our findings confirm that while using larger image sizes (especially for 2D datasets) yields slightly better performance, competitive CBMIR performance can still be achieved even with smaller image
  arXiv  Detail & Related papers  (2024-09-14T13:07:30Z)
• MTP: Advancing Remote Sensing Foundation Model via Multi-Task Pretraining [73.81862342673894]
  Foundation models have reshaped the landscape of Remote Sensing (RS) by enhancing various image interpretation tasks. transferring the pretrained models to downstream tasks may encounter task discrepancy due to their formulation of pretraining as image classification or object discrimination tasks. We conduct multi-task supervised pretraining on the SAMRS dataset, encompassing semantic segmentation, instance segmentation, and rotated object detection. Our models are finetuned on various RS downstream tasks, such as scene classification, horizontal and rotated object detection, semantic segmentation, and change detection.
  arXiv  Detail & Related papers  (2024-03-20T09:17:22Z)
• Towards a clinically accessible radiology foundation model: open-access and lightweight, with automated evaluation [113.5002649181103]
  Training open-source small multimodal models (SMMs) to bridge competency gaps for unmet clinical needs in radiology. For training, we assemble a large dataset of over 697 thousand radiology image-text pairs. For evaluation, we propose CheXprompt, a GPT-4-based metric for factuality evaluation, and demonstrate its parity with expert evaluation. The inference of LlaVA-Rad is fast and can be performed on a single V100 GPU in private settings, offering a promising state-of-the-art tool for real-world clinical applications.
  arXiv  Detail & Related papers  (2024-03-12T18:12:02Z)
• Confidence-Aware RGB-D Face Recognition via Virtual Depth Synthesis [48.59382455101753]
  2D face recognition encounters challenges in unconstrained environments due to varying illumination, occlusion, and pose. Recent studies focus on RGB-D face recognition to improve robustness by incorporating depth information. In this work, we first construct a diverse depth dataset generated by 3D Morphable Models for depth model pre-training. Then, we propose a domain-independent pre-training framework that utilizes readily available pre-trained RGB and depth models to separately perform face recognition without needing additional paired data for retraining.
  arXiv  Detail & Related papers  (2024-03-11T09:12:24Z)
• Egocentric RGB+Depth Action Recognition in Industry-Like Settings [50.38638300332429]
  Our work focuses on recognizing actions from egocentric RGB and Depth modalities in an industry-like environment. Our framework is based on the 3D Video SWIN Transformer to encode both RGB and Depth modalities effectively. Our method also secured first place at the multimodal action recognition challenge at ICIAP 2023.
  arXiv  Detail & Related papers  (2023-09-25T08:56:22Z)
• Interpretable 2D Vision Models for 3D Medical Images [47.75089895500738]
  This study proposes a simple approach of adapting 2D networks with an intermediate feature representation for processing 3D images. We show on all 3D MedMNIST datasets as benchmark and two real-world datasets consisting of several hundred high-resolution CT or MRI scans that our approach performs on par with existing methods.
  arXiv  Detail & Related papers  (2023-07-13T08:27:09Z)
• LVM-Med: Learning Large-Scale Self-Supervised Vision Models for Medical Imaging via Second-order Graph Matching [59.01894976615714]
  We introduce LVM-Med, the first family of deep networks trained on large-scale medical datasets. We have collected approximately 1.3 million medical images from 55 publicly available datasets. LVM-Med empirically outperforms a number of state-of-the-art supervised, self-supervised, and foundation models.
  arXiv  Detail & Related papers  (2023-06-20T22:21:34Z)
• UViM: A Unified Modeling Approach for Vision with Learned Guiding Codes [91.24112204588353]
  We introduce UViM, a unified approach capable of modeling a wide range of computer vision tasks. In contrast to previous models, UViM has the same functional form for all tasks. We demonstrate the effectiveness of UViM on three diverse and challenging vision tasks.
  arXiv  Detail & Related papers  (2022-05-20T17:47:59Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.