The NIC has a bug

[kyrieb-ekat]

Discovered in #1235 wherein troubleshooting/bug hunting has revealed that, for some specific images, something about image edges presence (we think) cause the job to get stuck on processing and never progress to "finished" or "failed". Making a separate issue for addressing/unpacking the NIC specifically, with the debugging steps and examples describes in #1235 .

It is possible to do a full e2e run with images which have previously failed by cropping the images so that only the page being OMR'd is visible, so this is nothing blocking or breaking, but is something to keep in mind for future e2e OMR runs.

[homework36]

Ater testing rodan-main/code/rodan/jobs/gamera_rodan/wrappers/classification.py at around line 103, I believe this is an issue inside Gamera 4 classify.py. The recursive method _classify_list_automatic is not going to terminate itself properly given some inputs (possibly with the edge)

[fujinaga]

That would make perfect sense. Will you be able to fix it? Ask co-pilot for assistance.

[homework36]

Should I fix it directly in gamera4 repo or the gamera wrapper in Rodan?

[fujinaga]

Whichever is easier for you, then create an issue to fix the other one.

[kyrieb-ekat]

Gamera4 will require fewer hoops to jump through than getting gamera (OG) going locally; if you do start with Gamera both Lucas and I got Gamera successfully working locally this past summer if you need any assistance/insight into what we did.

[homework36]

My understanding is that if we fix gamera 4 then it's better in general and we might not need to update the wrapper in Rodan. However, that's more complicated for sure and I need some time to play around with it, too.
@kyrieb-ekat Yes I will definitely need your expertise! Thank you in advance!