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/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

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

Zendy Tools

Zendy Open

Surgical phase recognition is of particular interest to computer assistedsurgery systems, in which the goal is to predict what phase is occurring ateach frame for a surgery video. Networks with multi-stage architecture havebeen widely applied in many computer vision tasks with rich patterns, where apredictor stage first outputs initial predictions and an additional refinementstage operates on the initial predictions to perform further refinement.Existing works show that surgical video contents are well ordered and containrich temporal patterns, making the multi-stage architecture well suited for thesurgical phase recognition task. However, we observe that when simply applyingthe multi-stage architecture to the surgical phase recognition task, theend-to-end training manner will make the refinement ability fall short of itswishes. To address the problem, we propose a new non end-to-end trainingstrategy and explore different designs of multi-stage architecture for surgicalphase recognition task. For the non end-to-end training strategy, therefinement stage is trained separately with proposed two types of disturbedsequences. Meanwhile, we evaluate three different choices of refinement modelsto show that our analysis and solution are robust to the choices of specificmulti-stage models. We conduct experiments on two public benchmarks, theM2CAI16 Workflow Challenge, and the Cholec80 dataset. Results show thatmulti-stage architecture trained with our strategy largely boosts theperformance of the current state-of-the-art single-stage model. Code isavailable at \url{https://github.com/ChinaYi/casual_tcn}.

Not End-to-End: Explore Multi-Stage Architecture for Online Surgical Phase Recognition