Wrong video is being dropped from visualizer when pose estimation confidence is low

[mattpetrucci]

User reports that a video drops in the visualizer, and it is due to low pose estimation confidence. It looks like the wrong video is dropped.

https://simtk.org/plugins/phpBB/viewtopicPhpbb.php?f=2385&t=23905&p=55521&start=0&view=

[carmichaelong]

It looks like the wrong video is dropped.

I want to clarify "wrong video is dropped". This is what I suspect but haven't 100% confirmed: Based on some similar examples I've seen, the processing likely correctly "drops" the video with low confidence, but since the videos are jumbled in the folders, the "wrong video" is shown in the visualizer.

[Seeeeeyo]

It looks like the wrong video is dropped.

I want to clarify "wrong video is dropped". This is what I suspect but haven't 100% confirmed: Based on some similar examples I've seen, the processing likely correctly "drops" the video with low confidence, but since the videos are jumbled in the folders, the "wrong video" is shown in the visualizer.

I think I fixed that. I encountered the same issue, and it is now fixed for the trials I had the issue with. Please check the trials users reported to double check

[mattpetrucci]

@Seeeeeyo Do you know what PR you fixed this in? Was the problem what Carmichael mentioned above?

[Seeeeeyo]

@Seeeeeyo Do you know what PR you fixed this in? Was the problem what Carmichael mentioned above?

7716ecb

[mattpetrucci]

Thanks @Seeeeeyo. I checked one of the trials that they mentioned and it looks like the right videos are in the visualizer (the 2 of 3 cameras that had a sync video).

@Seeeeeyo Can you confirm the root cause just so we have it documented? It looks like per your PR that we just needed to explicitly index the videos vs. trusting data.videos had the right trial info.

And @carmichaelong is that consistent with your reasoning above. Thanks!

[carmichaelong]

Thanks @Seeeeeyo! This fix makes more sense than my reasoning above (since there's no folder structure in the bucket). We'll have to see if this has downstream effects on fixing opencap-org/opencap-core#228 or if the functions for downloading and zipping files there still needs to be tweaked.

For documenting, one place this could be helpful to do is use CHANGELOG.md for even small user-facing issues like this so we can trace. And then we could pick specific ones that particularly affect users and add a running, dated shorter version on the web app too.

[Seeeeeyo]

Thanks @Seeeeeyo! This fix makes more sense than my reasoning above (since there's no folder structure in the bucket). We'll have to see if this has downstream effects on fixing opencap-org/opencap-core#228 or if the functions for downloading and zipping files there still needs to be tweaked.

For documenting, one place this could be helpful to do is use CHANGELOG.md for even small user-facing issues like this so we can trace. And then we could pick specific ones that particularly affect users and add a running, dated shorter version on the web app too.

Sounds good, we can use the changelog to document, works for me