Search papers, labs, and topics across Lattice.
The paper introduces SurfSurg6D, a dense-correspondence framework for 6D pose estimation of textureless surgical instruments, addressing challenges like limited data and lack of texture. They generate a synthetic dataset, SynSurg6D, to augment existing datasets and diversify pose distributions. SurfSurg6D leverages RGB images to establish dense correspondences between the image and a 3D instrument model, achieving state-of-the-art performance on SurgRIPE, EndoVis2018, and SurgPose datasets.
Synthetic data can overcome data scarcity and textureless challenges to enable surprisingly accurate surgical instrument pose estimation from RGB images alone.
Surgical instrument pose estimation provides crucial information for promising applications, including autonomous robotic surgery, skill assessment, and standardization of surgical workflow. However, this task remains highly challenging due to high precision requirements, frequent occlusions, textureless instruments, scarcity of depth information and very limited annotated data. These constraints often lead to unsatisfactory performance when employing general object pose estimation approaches to surgical scenarios. To address these issues, we first construct a new dataset SynSurg6D, to alleviate the data shortage in this task. We further propose SurfSurg6D, a dense-correspondence framework tailored for surgical instrument pose estimation. Experimental results on the SurgRIPE, EndoVis2018 and SurgPose datasets demonstrate that the introduction of our generated dataset SynSurg6D is able to diversify the pose distributions, thus enhancing the performance of existing approaches. Furthermore, SurfSurg6D outperforms existing methods, providing a robust solution for precise and efficient RGB-only pose estimation.
