Boosting Generalizable Depth Estimation in Endoscopy by Mixture of Lightweight Experts and Intrinsic Image Alignment
This work addresses the challenge of generalizable depth estimation in endoscopic surgery, which is critical for 3D perception in medical robotics, but the gains are incremental over existing methods.
The paper introduces EndoMINI, a self-supervised framework for depth estimation in endoscopic scenes, which uses a mixture of low-rank experts for parameter-efficient fine-tuning and an intrinsic image alignment loss to handle illumination interference. The method achieves state-of-the-art performance on SCARED, Hamlyn, and SERV-CT datasets, with improvements in depth and ego-motion estimation.
Depth estimation is a significant task for 3D perception in endoscopic surgeries. However, illumination interference and feature diversity in various endoscopic scenes are still challenges for generalizable depth estimation and ego-motion estimation. Based on this, a novel self-supervised framework, EndoMINI, is proposed for depth estimation in endoscopic scenes. Specifically, mixture of low-rank experts (MiLoRE) is proposed to perform parameter-efficient fine-tuning, which can also boost the model adaptation to scenes with different characteristics. Meanwhile, an intrinsic image alignment (IIA) is introduced into the training loss to alleviate the influence of light reflectance in endoscopy with a novel intrinsic image decomposition network. The proposed method is evaluated on SCARED datasets for supervised depth estimation, and two endoscopic datasets, Hamlyn and SERV-CT, for zero-shot depth estimation, compared with state-of-the-art works as well. The experimental results demonstrate outstanding performance of the proposed model and the effects of the main contributions.