A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is
Segmentation for tracking surgical instruments plays an important role in robot-assisted surgery. ... In this paper, a novel network, Refined Attention Segmentation Network, is proposed to simultaneously segment surgical instruments and identify their categories. ... To overcome the issues mentioned above, a novel network, Refined Attention Segmentation Network(RASNet), is proposed which utilizes attention mechanism for semantic segmentation of surgical instruments ...arXiv:1905.08663v2 fatcat:ec22cobptrblvooyb3lt53oosi
To tackle the issue of real-time surgical instrument segmentation for more precise instrument tip localization, we propose the YOLOv3 and ResNet Combined Neural Network. ... The real-time prediction of next desired camera location is estimated using segmented instrument's tip locations from endoscope video and surgeon's attention focus given by tracked virtual reality headset ... The current state-of-the-art method Refined Attention Segmentation Network (RASNet), combines ResNet-50 pre-trained on ImageNet as encoder and Attention Fusion Module (AFM)  . ...doi:10.1109/access.2021.3079427 fatcat:ac7wkfruirdgvnkpayqlunkpqy
While numerous methods for detecting, segmenting and tracking of medical instruments based on endoscopic video images have been proposed in the literature, key limitations remain to be addressed: Firstly ... Intraoperative tracking of laparoscopic instruments is often a prerequisite for computer and robotic-assisted interventions. ... Furthermore, the authors wish to thank Tim Adler, Janek Gröhl, Alexander Seitel and Minu Dietlinde Tizabi for proofreading the paper. ...arXiv:2003.10299v2 fatcat:bghqxe7rsbeklhu3nqe6poudxm
Accurate and robust scale estimation in visual object tracking is a challenging task. ... First, the location and scale of the target object are predicted in an anchor-free fashion by decomposing tracking into parallel classification and regression problems. ... Acknowledgments: We gratefully acknowledge the support of the NVIDIA Corporation with the donation of the Titan V GPU used for this research. ...doi:10.3390/s22010354 pmid:35009905 pmcid:PMC8749605 fatcat:qbz6sxf3f5anbkphpjrjot6m7a