Table 2 Performance of the STALNet model on Surgical Phases across different encoders.

From: TEMSET-24K: Densely Annotated Dataset for Indexing Multipart Endoscopic Videos using Surgical Timeline Segmentation

Phase Name

ConvNeXt

ViT

SWIN V2

Accuracy

F1 Score

Accuracy

F1 Score

Accuracy

F1 Score

[01] Setup

0.99 ± 0.09

0.97 ± 0.02

0.98 ± 0.13

0.94 ± 0.05

0.99 ± 0.10

0.97 ± 0.03

[02] Dissection

0.99 ± 0.10

0.99 ± 0.00

0.97 ± 0.17

0.97 ± 0.00

0.99 ± 0.11

0.99 ± 0.00

[03] Specimen Removal

1.00 ± 0.03

0.97 ± 0.03

1.00 ± 0.04

0.95 ± 0.05

1.00 ± 0.02

0.99 ± 0.01

[04] Closure

0.99 ± 0.09

0.99 ± 0.00

0.98 ± 0.14

0.98 ± 0.01

0.99 ± 0.08

0.99 ± 0.00

[05] Scope Removal

1.00 ± 0.03

1.00 ± 0.00

1.00 ± 0.05

0.99 ± 0.01

1.00 ± 0.03

1.00 ± 0.00