Closed brian-c closed 10 years ago
I'm somewhat against making any drastic changes to enable the classification summary work any better than it does right now. Feels like a rabbit hole where we could spend a lot of time trying to improve the results without getting very far.
That said, I would be for putting a message at the top of the summary, saying something like "these are a best-guess of the condors you marked, and should not be taken as confirmation that you marked correctly or incorrectly." rather than "X confirmed condors".
I like Chris's idea.
I agree, that seems like a good fix!
I guess I don't understand why we can address the location part of this, which is causing the majority of missed IDs. We included the Datasource column in the Tag table because it was required to ID the condors. Isn't there a way to program in photo source, and then apply the tag table based on photo source (Socal vs other)? The Bios will be fairly meaningless unless the Socal data are split off, and the users really like the Bios and read them, but are currently frustrated. I think this would result in a big improvement in IDs and shouldn't be a rabbit hole.
I guess more importantly, we'll need to get correspondence between the IDs and the tags from photos for any of our analyses to make sense.
Are "Socal" condors the only special case?
We have a directory path for each photo. Can we reliably determine the photo's site from that path? I'm talking about something in a format like this:
USFWS photos/Remote Feeding Site Photos/Remote Feeding Photos_2013/Bitter Creek/NRFS/NRFS Jan 5-12, 2013 = YES/PICT1098.JPG
When we come across a photo from a Socal site (determined by the directory path), we can use that to search for Socal condors.
Am I on the right track here?
That's right. The Socal birds don't mix with the birds from the other sites so they haven't coordinated their tags to avoid overlap. USFWS manages all Socal sites and only the Socal sites. That should work!
I've also seen a few reported misIDs arising when the Pattern column is NA.
Just deployed a change that checks that the string "USFWS" exists in the photo's original path, and if so limits the possible condors only to the Socal-sourced ones. Feel free to close this issue if you can verify that it's lead to fewer misidentifications.
Yeah! Thank you! Fingers crossed
So the frequency of misID complaints seems to have dropped, but there are still puzzling problems: Here are some examples:
Ugh, sorry. A typo made all sites identified as Socal. Should be fixed now.
@vjbakker Any idea who this guy is? http://talk.condorwatch.org/#/subjects/ACW0003yf7
I'm not seeing any "9"s or black tags with two dots in the data.
In fact it looks like there are no black tags with two dots at all. Here's another: http://talk.condorwatch.org/#/subjects/ACW00040is
Well, yes, those are the remaining problems that I've been alluding to. We found out last week that in the early years of photos (2006-2008), Ventana Wilderness Society had a different tag system. We just sent another email today to try to get the info from them. They are stretched thin and we may need to send someone out to dig through their files. For now, we do not know. Would be great not to post them but I suppose it's too late. BTW, I'm so thankful you are on this Brian!
Yeah, I think we can close this, bearing in mind that there is some older tag data we need to add at some point..... Except I don't know how to close it....
Great to hear it worked out in the end. Happy to close it for you!
Some live condors have the same tag others at the same time.
Check subject timestamp vs condor birth and death?
Location matters too. No idea how to fix this.