Guidelines on minimum Observation Standards

Hello iNat Community,

I know that the data quality issues was discussed in regards of first time users that upload poor quality photos or such. With this thread I do not want to discuss about the photographic abilities of the users.

My concern is about, what should actually be considered as the minimum standard for an observation. I came across a lot of photographs of images shown on screens, where users had taken a snapshot from a camera display, computer screen, or even another smartphone screen showing an image. Or for Bat species, many users just make a screen shot of a spectrogram or a bat detection software. I have also seen screen shots of other ID Apps like BirdNET. I think this is problematic.

In my opinion an observation should be original, first hand and be confirmable by the community. What do you think?

Originality
I think the original data recorded of the species should be part of the observation data. So data transformations should not be the only evidence. A screen shot of a spectrogram is not the original evidence, it is a visual representation of an audio, processed with a fourier transformation. It can be necessary for bat identification but without the original audio it is not reproducible. There is also no standard for spectrograms, they are arbitrary in coloring and so on, a lot of times you can not even read the axis legends (time and frequency domain) because of poor image quality. Same goes for screen shots form other apps like BirdNet.
A lot of people upload slightly altered photographs, with contrast correction and so on. I think this is ok when it does not change the actual appearance of the species.

First hand
I think a photograph of a screen should not count as an observation at all. If its your camera display you are photographing, you could just upload the images when read from Memory Card. If its not your own image, its not your own observation. This is like observation from hearsay. And also it is not original. Scanned old but recent analog images would be ok, as the digitization is needed to upload.

Confirmability
I think the observation should be confirmable from the community. This point is most controversial I guess, as we can not always be sure about all aspects that could lead to an identification. But some times there are pictures where its not even clear what in the image is the subject, e.g. where the user says the observation is a caterpillar but photographed from like 3 meters away on the forest floor.

To address this issues, I think there could be more options in the data quality assessment pane to deal with this. One solution could be, to introduce a guideline on the minimum standards and give a button “does not meet standards” or such. Or a thumbs up / down for each image like “this is not helping for ID”

But however this is dealt with, I think there should be some guidelines on the minium standard for an observation.
What do you think?

5 Likes

I mark these as “no evidence of organism” because the only thing you proved is that there was sound- Never mind, I guess it IS possible to read a spectrogram to species sometimes

I disagree since your phone captures location and some cameras don’t. As for photos of screenshots, I usually give the benefit of the doubt as long as the observation location isn’t close to or inside of a building

That’s already a requirement. If it can’t be ID’d, then it can flagged as “(community taxon) is as good as it can be” then it will be put into whichever grade fits (family or lower=research and superfamily or higher=casual, iirc)

5 Likes

I think marking spectrograms “no evidence of organism” just because you don’t like this as evidence is simply vandalism and you should not do that. There is no standard way how to handle ultrasound on iNat - surely you can upload a wav, but for some people that’s simply too technical to do and for others it’s not useful anyway, because to interpret it, you need specialized software. Yes, a random screenshot is usually not enough for bat ID - but sometimes it is, at least to genus, depending on what was included in the screenshot.

And while I find the screenshots of back of camera etc. a bit funny, you have to understand that for some people getting files from camera to iNat is difficult. It may even be hard to comprehend HOW difficult - yes, for me the concept of a file that I havea complete control about and can copy and edit as I please, is natural, but a lot of stuff today works like “this app exports this to that app” and when it doesn’t walk like that, some users are truly lost.

I think the goal of iNat is to facilitate connection of people with nature, not to collect super high quality data, so no, we don’t want to create MORE barriers for people to post observations.

27 Likes

I disagree since your phone captures location and some cameras don’t. As for photos of screenshots, I usually give the benefit of the doubt as long as the observation location isn’t close to or inside of a building

At least some of these are placeholders too. I’ve done things like posted observations from my phone that are really terrible quality, because I want the location data saved until I have a chance to upload from my computer. I try to make a note or wait to upload until just before I’m going through the camera pictures.

17 Likes

Back of the camera - I leave a comment, but I won’t ID. (Please upload the original picture)

Identifiers need enough field marks to add an ID. If I can’t see them, next …

I also choose not to ID any sound clips.

We make it easy for newbies to upload obs. But if you have hundreds, or even thousands of obs - then yes minimum standards which rise with the volume of obs you upload. (How often must we remind you - pretty flower, what do the leaves look like, is it a tree or a littl’un??)

I’ve never seen the problem with photos of photos as evidence. It’s no more or less “first hand” evidence than a scanned analog photo or a photo that’s been lightly edited/stacked. I get that in most cases the observer probably could upload a better version of the picture, but if they post a picture of a picture of a robin they just saw, I have a hard time saying that’s not evidence that they just saw a robin; unless you think they’re lying about having taken the picture. It’s sure more of a “first hand” interaction with the organism than trail cam photos, and those are considered fine to post. The main concern I hear about these cases is “how do you know they really took the original picture?” but that’s a question one can ask about every photo upload- whether you post it as a JPEG downloaded from an SD card, a scan of a printout, a photo of a printout, a photo of a screen, a screenshot, or whatever… I don’t see how a photo of a back of a camera is inherently any more or less reliable than any other file uploaded. And as was mentioned above, a phone shot of the photo just taken on the camera is more likely to have a precisely accurate location tied to the file than the camera file itself would have.

The only gray area I see is the copyright issue- depending upon how much of the camera display or phone app is visible, posting it may or may not have copyright repercussions. But I know that’s already had its own thread on here before.

As for the general idea of confirmanility, I think it’s hard to make any specific standards about how that relates to photo quality. I’ve seen some absolutely abysmal looking photos that taxon experts can still confidently take to species level or below, and also some super closeup fancy macro images that can’t be identified past family because particular features are hidden. So while there’s certainly a trend present where better quality photos are more identifiable, I would never want to discourage someone from posting just because they think the picture is bad. Unless I’m personally an expert in that taxon, I don’t claim to know what makes a photo of that taxon identifiable.

16 Likes

I think photos of spectrograms only prove “it’s loud,” they don’t actually provide any evidence of what organism was calling or even if there was an organism calling. I don’t do it because “I don’t like that;” I do it because I literally believe it isn’t evidence.

Can it actually be deduced that something was calling from a spectrogram? Because I’ll change my mind if so

1 Like

As for the general idea of confirmanility, I think it’s hard to make any specific standards about how that relates to photo quality. I’ve seen some absolutely abysmal looking photos that taxon experts can still confidently take to species level or below, and also some super closeup fancy macro images that can’t be identified past family because particular features are hidden. So while there’s certainly a trend present where better quality photos are more identifiable, I would never want to discourage someone from posting just because they think the picture is bad. Unless I’m personally an expert in that taxon, I don’t claim to know what makes a photo of that taxon identifiable.

One of the examples that comes to mind for me if a picture of what was presumably a bird that was one of the blurriest I’ve seen, and I wondered why the user had even bothered to post it- that was, until the very next second where I realized “oh, sure, well that’s a male American redstart alright” and I am far from an expert.

It’s absolutely bonkers from outside perspectives to try and figure out how much, or what sort of information is needed to identify something. I work at a nature center, and a bit later this summer I’ll have visitors start asking “hey, what’s the plant with the really big leaves…” and that’s all I need them to say before answering that it’s cow parsnip, because I saw what part of the trail they were coming from, and know what is most prevalent there, and I’ve gotten the question a thousand times before. And the same guests might describe something else they saw with a ton more detail, and I’ll have no idea what they’re talking about, which is obviously confusing from their perspective :sweat_smile:

12 Likes

I have a fair number of objectively bad photos on iNat. I post them if I feel that there is a chance that somebody who knows something will be able to see what I saw, not because they’re going to win any prizes. If the response is “jeez, that really is a useless photo”, so be it. That’s what the community ID process is for.

I also play with contrast, saturation and image brightness if I think it will improve the image. I used to let software deal with that autonomously but realized that it was messing with colours in dubious ways and went back to manually editing. My rule is that information captured by the camera can be used to present the most usable image even if the camera didn’t use the information in the best way for that purpose when it produced the image. There is no reason to do otherwise if the objective is to present an image that matched what was observed rather than an image that matches the camera settings.

I have recently started collecting audio observations, partly because I have become more confident about birding by ear and partly because it causes me to pay attention to different things when I’m poking about in the bush. Some of my early audio observations weren’t very good at all, but I learned by repetition. Learning is iNat’s raison d’être. My most recent audio observation is a groundhog, which I was not expecting.

11 Likes

I don’t understand this argument at all. I can ID many bird species just from looking at their spectrograms - my hearing is not great, so I actually nowadays “listen” to birds like that - because this gives me the high notes that I can’t hear. I for example found out that European Robin, one of the most common birds here with a super recognizable song and a call that.I have known for a long time has another distinct call that’s above my hearing threshold - but I can now clearly tell it from t he spectrogram, because it’s surprisingly unique.

15 Likes

I agree there should be no more barriers for people to post, but there should be more or better ways for quality control of data by the community. iNat has also the aspiration to provide useful data for scientists. Its one of the foremost reason for me personally to post observations in the first place. This does not mean data should be deleted or something, just better labeled like with the non-wild animals.

Thats exacly my point, we should have a discussion how to handle this. Including a picture (screenshot or photograph of screen) of a spectrogam will just mess up the photo identification AI in the end, as it is a totally different domain of information, when mixed with all the actual photographs of bats. Maybe that could be a feature request to create a standardized spectrogram through iNat when sound is uploaded. This way everyone would be on the same level for identification by spectrogram.

I get your points, that the picture of a screen is in most cases enough to ID most species and is more convienient to take a snapshot than to upload a original picture to handle location and easier than copying form memory card to pc for upload. But it is still a very different thing, with less quality for research, e.g. developing the AI or other image based analysis. It should be no problem for common species, but for rarely observed species it is a loss for the research community. Also I would still argue that it is problematic in case of secondary information and potential copy right issues.

Me too :) Thats nothing I wanted to blame or discuss, it could be still valuable information when the key features are seen for ID. I just thought about it would be a good idea to have a better way to mark obviously non ID able pictures as a form of quality label.

If individuals have not been on the forum much, it can be hard for them to understand exactly how helpful people like you, and others on the forum, can be. I continue to be mind-blown by how helpful people on the forum are, even when I am asking questions that have been dealt with before (but I couldn’t find where), even when my questions seem to be to be elementary and, dare I say it?, stupid. On the observations themselves, only rarely. But on the forums, so generous!

1 Like

Yes, I often have no clue, and I’m the one taking the pictures. I treasure the random comments that make me less ignorant, e.g., you really need a head shot for caterpillars, you need a close up of genitalia for that insect, or you will never get it to species without genitalia dissection.

4 Likes

Kindred spirits.

3 Likes

Yes it is possible to ID bat calls by interpreting the spectrogram. But to be useful, a spectrogram, like any other diagram needs to be clear in what it shows and how it was made. This is mostly not given with the screen shots I have seen. On Y axis you typically should have the frequency in Hz on X axis you typically have the time in seconds. The plotted spectrum is a time variant of the frequency distribution for a certain time step which was typically produced with a fourier transformation. There are several ways to process a spectrogam and thus their appearance (what you want to interprete) can change dramatically, e.g. by the time step intervals you choose or the frequency range or even the aspect ratio of your plotting area. So your argument is still mostly valid. I would agree when the screen shot is the only evidence given and no audio is attached.

2 Likes

I think the currently available tools cover what you’re asking for.

iNaturalist allows you to add an ID to the taxonomic level that the evidence supports, and there’s an entire Data Quality Assessment section on each observation that allows any member of the community to weigh in on various aspects of the observation’s data, including whether it can be IDed finer than the current community ID. I don’t think we need more of these tools or more arbitrary guidelines than what are are already there. The community can vote and discuss on the observation.

As for photos of screens: if it’s from the back of the camera that’s not great but doesn’t cross the threshold of being lack of evidence. A photo of a computer screen may be evidence of copyright infringement, but that’s not always the case. Many people don’t know they can post photos from a computer via the website, so they use the app. So you can write a comment and ask the observer if it’s their photo.

22 Likes

When I ask taxon specialists for help on obs, they are almost always helpful. I have to learn to accept that even the taxon specialist cannot ID That One. (But I am glad to see the last of the CNC residue 6K IDs later)

3 Likes

I don’t know what was so complicated about my “”“argument;”" all I asked was if this was possible because I was told it wasn’t before

At least some of these are placeholders too. I’ve done things like posted observations from my phone that are really terrible quality, because I want the location data saved until I have a chance to upload from my computer.

I do this as well. As far as spectrograms, I think that’s mainly useful for IDing bats at night. It’s virtually impossible to photograph flying bats at night, but you can record them and actually ID them (at least to genus and sometimes species) based on the ultrasonic sounds they make. You could post the actual sound, but humans can’t hear it, so it would just sound like silence. So some people slow down the sound recordings and post that and other people post the spectrograms.

6 Likes

“I’ve done things like posted observations from my phone that are really terrible quality, because I want the location data saved until I have a chance to upload from my computer”

Perhaps you could use a copypasta - which you could delete later - for identifiers - please don’t ID yet, will upload a better picture later
That is also a situation where the requested draft mode would be a better solution all round.
https://forum.inaturalist.org/t/create-a-draft-mode-for-uploaded-observations/2538

1 Like