Data Collection

Data Collection  

  By: jcwang on Aug. 8, 2022, 3:21 a.m.

Hi, we have two questions about the data collection that:

1) whether the clips of training dataset belong to the same or unique patients? 2) whether the clips of final evaluation will be collected from totally different patients?

Thanks!

Re: Data Collection  

  By: aneeqzia_isi on Aug. 9, 2022, 4:44 p.m.

Hi jcwang,

All the videos are collected from robotic training cases primarily performed on porcine models (no patient involved). The clips can be from the same and different training cases as one training can be very long (~3-4 hrs). The way training and testing data are collected are the same - so both sets will have clips from same and different robotic training cases.

Best, SurgToolLoc 2022 Organizing Committee

Re: Data Collection  

  By: jcwang on Aug. 10, 2022, 5:33 a.m.

Thanks for your reply! There is one more question about the test metrics. As the data amounts of different classes are extremely imbalanced, whether the "average f1 score" is calculated through (a) straightly averaging all classses, or (b) weighting each classs.

Re: Data Collection  

  By: aneeqzia_isi on Aug. 12, 2022, 6:02 a.m.

Hi jcwang,

There is no weighting in the average f1-score metric and is calculated directly as an average.

Best, SurgToolLoc 2022 Organizing Committee