We have a few questions regarding the training data annotations:
1. For frames where a step is not specified in the annotations file, the step of these frames corresponds to step 0?
2. Since the training data does not include frames whose label corresponds to step 10, are our models not expected to predict this step in the validation and test stages?
Thanks in advance,
Created by Cristina González Osorio ci.gonzalez10 Yes, as you mentioned, each frame has a label correspond to it and all phases are included in the annotations. The issue was that the editor we were using changed the original format of the csv files.
Thanks for your feedback,
Best,
Cristina Can you please confirm that our answers were clear enough and handle the issues you pointed out?
Thanks.
Thanks for your interest in our challenge.
Please find our answers below:
1. Unless it is a mistake from our side, each frame should have a label correspond to it. Can you please point out to the annotation files that have missing data?
2. Step 10 (Irrigation/Aspiration) exists in most of the videos. If you mean Step 9 (Vitrectomy), it exists in train19. We have modified (after the 1st of April) the list of steps that were annotated. You can check them again in the [Data](https://www.synapse.org/#!Synapse:syn21680292/wiki/601563) section.
Regards,
Hassan,