Better Image Matches: Photo Similarity Update to Computer Vision Suggestions

In case you missed the announcement on the website, we just rolled out new functionality for Computer Vision ID suggestions! Suggestions will now display the taxon photo that’s most visually similar to the observation photo submitted to Computer Vision.

For example, if the default photo for a moth species depicts the adult moth but you submit a larva photo of that species, the ID suggestion will display a larva photo if available.

You can read more about it in this blog post.

18 Likes

It’s a feature some of us never realized we wanted!

12 Likes

I was thinking the same thing when I saw the blog post. I think this will be a very useful feature.

4 Likes

Extremely helpful feature. Have been hoping for something like this for a long time.
But I didn’t even think of the little sugesstion photo UT the species page itself. Because if you had different life stages you often would have to go into the images for this species and filter for example for the larval stage which took a bit of effort to do.
Displaying visually similar photos on these two is a huge time saver.

3 Likes

Great improvement. Thanks! Especially, where you are not an “expert” in the species you are observing, this is much better helpfull than before.

1 Like

How that all taxon photos are a lot more relevant, could you please unlock the photos for a lot of the taxa that have them locked?

is the similar photo coming only from the set of up to 12 taxon photos? i see some photos in the computer vision results that look like they might be coming from elsewhere.

for example, here’s the suggestion that i see on one of my Loggerhead Shrike observations. it’s a bird flying to the left with mouth open:

when i look at the 12 taxon photos associated with the taxon, i can’t find this photo anywhere:

and when i look to see if the taxon photos have been changed recently, i don’t see any changes since more than a month ago:

1 Like

There are a lot of subspecies for this species, so I would guess it is a taxon photo for one of the subspecies.

2 Likes

hmmm… it does looks like the similar photo is the photo for L l. ssp. mexicanus.

if it is coming from a subtaxon, i wonder if there would be a way to indicate which subtaxon it was coming from, in case i wanted to look at the specific observation where the specific photo came from?

3 Likes

Awesome! The lack of this annoyed me, but I never thought to make a feature request. Thank you!

2 Likes

Not sure if you read @pleary’s comment on the blog post, he said

the photos this feature is drawing from are photos of the target taxon plus all descendant taxa. Right now we can only open the taxon page to the suggestions-displayed photo if that photo is part of that suggested taxon’s own Taxon Photos. When the suggestions-displayed photo is associated with a descendant taxon and not associated with the suggested taxon, it will not be shown when clicking through to the taxon page.

This may be the case for higher-level taxa like we mention in the post (both the top-level higher-taxon suggestion as well as non-species model suggestions). This may also be the case when the suggested taxon is a species with infra-specific taxon descendants and the most similar photo was associated with one of the infra-specific taxa. We may be able to address this in the future. But for now, we are only able to feature the suggestions-displayed photo if the suggestion’s direct taxon has that photo as one of its own Taxon Photos.

I don’t think that answers your specific ask here, though.

2 Likes

I’ve been checking on the taxon photos a lot lately and noticed that the images that come up often are not within the 12 taxon photos listed on the taxon page. I assume this is a bug?

It is not.

See Tony’s comment above yours, mine above that, or the linked comments on the blog post.

4 Likes

I had this as well; the first species I tested the feature out on was Northern Cardinal. On an observation of a female cardinal, the image shown was a very low quality photo that wasn’t one of the species-level taxon photos. I assumed it was coming from one of the subspecies, but because there are like 20 subspecies I couldn’t be bothered to look through them all to find and fix it.

1 Like