When I uploaded a dragonfly photo today and used the auto ID feature, at first I was puzzled as to what was going on. The auto ID came up with a deer.
After I thought about this, I was wondering if this could be a software issue. The dragonfly is a “common whitetail”. The deer is a “white-tailed deer”. Did the auto ID correctly identify the dragonfly as a “common whitetail” and then did the code behind the auto ID get “white-tailed deer”?
Maybe I am overthinking this. But, as a former software tester I decided to enter it. When I used to test software the thinking was to report everything and let the programmers figure out whether or not it was an issue. In any case, this probably needs to be looked at. I don’t think the photo in any way resembles a deer.
Please move this post if this isn’t the correct forum sub category.
You can try uploading the photo yourself from my observation entry.
https://www.inaturalist.org/observations/81697069
Platform" Website
Browser" Chrome
What’s the name of your file? If it contains whitetail iNat uses it to auto fill the name and chooses deer.
3 Likes
@fffffffff nailed it. If I download the image, the file name is “original.jpg”. If I upload that, iNat guesses that it is indeed a dragonfly. If I rename the file to “whitetail.jpg” and upload that, iNat calls it a deer.
I guess this is arguably still a bug – why should “whitetail” in the name not find the dragonfly with that name as well as the deer with that name?
1 Like
I didn’t know that the iNat software considered the file name.
So, the file name “whitetail” probably triggered this.
Thank you for that. I will remember to name my files closer to a possible species in the future.
I usually allow for a generous amount of surrounding border space when taking a photo. There is nothing worse than cutting off part of a subject. I then crop a photo before uploading it. So, I change the file name when I crop and save a photo image file. I generally keep a copy of the original image file separately.
Is this something that should be posted somewhere in the “how to” instructions? It probably isn’t much of an issue if someone is out using a cell phone and not altering an image. But, it is obvious that a lot of people use cameras (non-phone cameras). So, chances are they are using some software to alter an original image and changing the original file name.
I did go back and change the name of the image file to “common_whitetail”. I uploaded it as a separate observation. The iNat software got the ID correct.
So, the software appears to consider the file name first. If it gets a match, it uses it. I know the website has to process a lot of images. But, the ID software should consider the image, as well. But, processing time/ability may be an issue with so many observations to deal with.
Computer vision is not enabled until you click on the area where the taxon name goes (when it has no ID yet). The process where it matched taxon names to file names is a separate system.
3 Likes
I clicked on that field which generated the list. See the list of deer? Unfortunately, the cursor is not captured in the screenshot. That list wouldn’t appear unless I clicked there, right?
In any case, now that I know that there is some software that is looking at the file name, I will be more specific when I rename a file.
Right - that list is a search result for taxa that match those words, “white-tailed deer”.
1 Like
