English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Instance segmentation of surgical instruments is a long-standing researchproblem, crucial for the development of many applications for computer-assistedsurgery. This problem is commonly tackled via fully-supervised training of deeplearning models, requiring expensive pixel-level annotations to train. In thiswork, we develop a framework for instance segmentation not relying on spatialannotations for training. Instead, our solution only requires binary toolmasks, obtainable using recent unsupervised approaches, and binary toolpresence labels, freely obtainable in robot-assisted surgery. Based on thebinary mask information, our solution learns to extract individual toolinstances from single frames, and to encode each instance into a compact vectorrepresentation, capturing its semantic features. Such representations guide theautomatic selection of a tiny number of instances (8 only in our experiments),displayed to a human operator for tool-type labelling. The gathered informationis finally used to match each training instance with a binary tool presencelabel, providing an effective supervision signal to train a tool instanceclassifier. We validate our framework on the EndoVis 2017 and 2018 segmentationdatasets. We provide results using binary masks obtained either by manualannotation or as predictions of an unsupervised binary segmentation model. Thelatter solution yields an instance segmentation approach completely free fromspatial annotations, outperforming several state-of-the-art fully-supervisedsegmentation approaches.

SAF-IS: a Spatial Annotation Free Framework for Instance Segmentation of Surgical Tools