Species ID upload bug

[Tony Wills]

There is a bug in the /upload page logic that sometimes saves the wrong ID, it is as though the system looks up what you've entered as the species again and selects one of the other options when it saves the observation after you hit 'submit'.

I first noticed this a while ago when uploading sea anemone observations, I drag and drop the photo on the upload screen, I then enter "Sea anemone" and choose "Sea Anemones (Order Actiniaria)", I then submit the observation.  I save the observation but invariably it shows my ID as "Sea Anemones and Corals (Class Anthozoa)".
Which is a little annoying but I wasn't too worried.

But now with http://www.inaturalist.org/observations/4528373, I **know** that I selected the seaweed, not the seal.  I have tried to repeat this, but instead of getting the seal I now end up with just a placeholder value entered, not the seaweed that I selected.

Attached are screenshots of my tests to show this isn't just finger trouble at my end :-)
(http://naturewatch.org.nz/observations/4539347, http://naturewatch.org.nz/observations/4539619)

(If it is of any help, I am using the Iceweasel browser v38.5.0 (Debian firefox clone), through the naturewatch.org.nz portal)

-- Tony

[Tony Wills]

Just did a follow-up test with a Windows 10 machine running a more up to date Firefox browser, same result.  But I did notice that if I downloaded the photo from iNat, and re-uploaded the same photo with drag an drop, it didn't change the entered ID (http://naturewatch.org.nz/observations/4540082).  The down/uploaded file has been resized and stripped of EXIF metadata etc by iNat, so not sure if that's the difference.

-- Tony

[James Bailey]

I wonder if this is related to the issue where sometimes clicking the "edit" button on an observation completely wipes the species and placeholder field.

Since I'll assume it isn't, I won't bring this issue into this topic.

[Tony Iwane]

Yeah, I wonder if it's a metadata thing? Did your original cystophora photo have "cystophora" as a keyword?

[Tony Wills]

No, I add no metadata to any of my photos.  Olympus spams all their photo files with "OLYMPUS DIGITAL CAMERA"  somewhere in the metadata, and the new upload page insists on grabbing it and putting it in the description, which I delete, but it often reappears! either while adding other photos, or after hitting submit.  ie I delete that text from the description only to see it back again.  I have to go and edit the observation.  I really, really wish it could be screened out (perhaps just ignore fields in all caps). (most of my photos are from an Olympus these days)

The wrong ID thing only happens with a few uploads, always with "Actiniaria" and now something funny with a few others (I uploaded something else yesterday, that turned my selection back into just a 'place holder').  One would hope that the manually entered fields would take priority over random interpretations of metadata, but there may be something in that.

-- Tony

[Tony Wills]

Just did a sea anemone upload with a non-olympus file (canon camera) same result (changed id to Anthozoa), but with a JPG without metadata, it uploaded ok.
http://naturewatch.org.nz/observations/4558255
http://naturewatch.org.nz/observations/4558311

-- Tony.

[Tony Iwane]

Tony, can you upload a photo that consistently causes this problem? I've tried with a few of my photos but haven't been able to replicate the issue.

As for the annoying OLYMPUS DIGITAL CAMERA thing, I'm able to get rid of it in Lightroom. What do you use to edit your photos?

[Tony Wills]

Adding any file that gets scaled down, because it exceeds iNat's 2048x2048 pixel max limit, and I enter as an ID of "Actiniaria" invokes the problem consistently.

I have uploaded a series of photos (each as a separate 'submit')
http://naturewatch.org.nz/observations/4571530 640x480
http://naturewatch.org.nz/observations/4571540 1920x1440
work ok.

http://naturewatch.org.nz/observations/4571548 3200x2400
http://naturewatch.org.nz/observations/4571533 4608x3456
show the bug

So repeatable with any file over the 2k limit.

As for the OLYMPUS thing, I'm sure I can edit it out, but that's not the point.  It obviously is a problem for anyone uploading unedited olympus photos, and there is also at least one other camera manufacturer with the same problem (I posted a bug report about that a while ago).  (and the ID bug happens with my cannon uploads too)

I will try and attach my smallest test which shows this bug (PB157663.JPG).

-- Tony

[Tony Wills]

I just downloaded the test file that I attached to my previous post, I see that google compressed the file and stripped some (all?) metadata but left the resolution intact (3200x2400).  I uploaded that and it still invokes the same bug, so strange as it seems the pixel width/height affects the taxon ID !??

-- Tony

[Tony Wills]

Hmmm, the copy I downloaded from this forum (as per previous post) only *sometimes* induces the bug. ie sometimes it retains the entered ID and sometimes changes to the other higher ID.  I have posted a copy of the original file on http://list.afraid.org/PB157663.JPG, which consistently shows the bug when I enter an ID of "sea anemone" and select the Actiniaria entry.

-- Tony

[Tony Iwane]

I tried that photo at both iNat and Naturewatch using Firefox on my Mac and can't replicate the error. Anyone else having a problem similar to Tony's?

[Tony Wills]

Interesting!  I just downloaded my own photo and uploaded it from a Windows 10 machine,
http://naturewatch.org.nz/observations/4572296
http://naturewatch.org.nz/observations/4572276

Firefox and MS Edge browsers.  Both show the problem.

Did you delete your test?  I can't see any under any of the three Tony Iwane accounts.  Could you upload the file, and point me to the test so I can compare other settings?

-- Tony

[Tony Iwane]

Here you go: http://www.inaturalist.org/observations/4572674

I will say, it was difficult for me to select the OLYMPUS DIGITAL CAMERA text to delete it when using Firefox as compared to Safari or Chrome...

On Saturday, November 12, 2016 at 1:58:44 AM UTC-8, Tony Wills wrote:

[Tony Iwane]

I think Ken-ichi was able to get to the root of the problem, Tony. Are you changing the taxon while the image is still uploading? Ken-ichi just wrote up an issue in GitHub about it.

https://github.com/inaturalist/inaturalist/issues/1194

On Saturday, November 12, 2016 at 1:58:44 AM UTC-8, Tony Wills wrote:

[Tony Wills]

Thanks, only differences I can see are the obvious ones of the timezone and description.  The resultant photos are identical.  I will see if I can get some others to try it too.  Might try changing timezone just in case, its a weird enough bug that who knows what might affect it.  Same bug on Chrome.  I haven't got a Mac to try it from, might try installing Safari and see what happens.

I am wondering whether it might be a bug that occasionally crops up for others, but they pass it off as having clicked on the wrong thing.

....edit...
I just looked at the http://naturewatch.org.nz/taxa/47533-Anthozoa page and noted it had "sea anemones" as a common name as well as "Sea Anemones and Corals" so I deleted that as it is rather superfluous.  Low and behold, the upload works ok now.  So it seems to me that the system was ignoring my selection of ID at some stage in the upload, and doing another lookup and finding the alternative taxon - and presumably the uploaders region settings effect exactly which common name was looked at. (attached screen shot of the entry that I deleted)

We could re-create that common name entry to try and reproduce the bug.  My original feedback post described similar problems with other taxons, but not as easily repeatable as this one.  Usually it just resulted in the right taxon name but as a 'place holder' rather than taxon entry.

-- Tony

-- Tony

[Tony Wills]

I am entering the taxon while the file is uploading if that's what you mean?

--Tony

[Tony Iwane]

Yes, that's what I was asking. That might be why you find the error in larger photos. Now there's a bug filed on GitHub, it's something the iNat team will be looking at. 

On Saturday, November 12, 2016 at 1:58:44 AM UTC-8, Tony Wills wrote:

[Tony Wills]

Thanks team for working the source of this one out so quickly :-)

I will delete all my test uploads, I presume they're no use to anyone now.

-- Tony.