Policy/recommendation about habitat photos?

Is there a policy or recommendations about posting habitat photos to the observations?

There’s one user in my country who posts lots of observations of bird sounds, and adds habitat photos to them. The taxon can be identified from the sound file, so the observations are valid for research grade.

However often all of the associated photos are about the habitat, and not the organism. Does this create problems to e.g. the computer vision (probably not, if there’s just few users doing this?), species pages, or other parts of iNat?

I would not like to forbid these kind of observations lightly: the user makes lots of valid observations, and seems to find iNat useful for their own purposes (e.g. mapping where urban birds live and collide with glass objects).

Examples:
https://www.inaturalist.org/observations/40142047
https://www.inaturalist.org/observations/39712613

4 Likes

Technically the organism should be in the photo and preferably fill a reasonable proportion of the frame. You could suggest to them that they host the photo offsite and then link it into the description using html image tags?

8 Likes

I think habitat photos can be extremely valuable source of data, so i’d rather figure out a way of handling them rather than getting rid of them.

Perhaps it would be possible to implement a ‘supporting photo’ tag for this and similar circumstances, where photos with such a tag would be exempt from the computer vision training data?

20 Likes

No one is suggesting getting rid of them :)

Only the first photo counts for the CV training, so this is usually not a problem as the context photo is typically placed after the whole frame organism photo. In this case the primary evidence is a soundfile, which would make the habitat (context) photo the first photo by default. This is incorrect, see comment by Cassi below…

iNat team are planning or considering the ability to mark individual photos with DQA, which will help with this problem, but I don’t know if it is going to be anytime soon.

In the mean time the inclusion of the photo by using html tags in the description and/or comments is a good way to include non-evidential images.

2 Likes

The photos taken for computer vision training aren’t only restricted to the first photo.

It is just the first photo that is used to compare against the trained model (when you click on the suggest an ID section).

All photos should depict the organism. The photos of organisms are included on taxon pages and hopefully there will be a way to view all photos in other areas of the site too: https://forum.inaturalist.org/t/ideas-for-a-revamped-explore-observations-search-page/8439/79

Agree with kiwi that photos that don’t depict the organisms should currently be put in the description, comments section, or in a journal post. iNat isn’t really set up to handle habitat photos, but feel free to submit a feature request! Maybe in an observation field redesign it could have a field type that allows uploading an image.

10 Likes

See also the related discussion at: https://forum.inaturalist.org/t/suggestion-on-how-to-include-habitat-images/3553?u=danaleeling
for a discussion on ways to include habitat photos in the description where they do not impact the computer vision training.

3 Likes

So I take it that there is no hard rule against adding habitat photos to observations, but a recommendation that this should not be done. (Please correct me if I’m wrong.)

I’m afraid that linking external photos to description is so difficult (especially with a mobile device), that most people will either continue adding habitat photos, or stop adding them altogether. This particular user would probably stop using iNat if someone would forbid adding such photos.

3 Likes

Provided that the organism is visible in the photo. The techniques in the related topic were done using the desktop website. A habitat photo that does not include the target organism would be problematic. There are computer vision training implications for images without target organisms.

5 Likes

It’s also good to remember that every photo included in an observation gets included in the collection of photos displayed for that species or taxon. We don’t want that collection to be cluttered by a bunch of photos that have no hint of the species in them. Personally I take a lot of habitat photos, and even before using iNaturalist, I always tried to make sure the subject species is visible in each one somewhere. To me that’s just practicing good documentation.

10 Likes

I understand that including a habitat photo can throw off the AI, but there are many times when a habitat photo can be useful when making an ID. Maybe someday a place for such a photo could be added?

4 Likes

If you do put in a feature request please post a link here so we can vote for it!

2 Likes

I think it would be very useful to have a place to add photos of habitats, host plants etc. related to the observation, but not showing the target organism. A good solution could be to allow for observation fields of type “photo”. Then, the habitat photos would be visible on the observation page but not used for the computer vision models. Would this be feasible? What do you think?

2 Likes

Why you want to add the photos to observations? Another website added them to places, but in the end this option and the photos disappeared.
https://www.inaturalist.org/places/search?utf8=✓&q=europe&commit=Zoeken

If you add an exact location to the photo and and exact location to the observation I expect you can link the observation with the habitat.

ps: no link to the feature request?

2 Likes