Agreed. If the subject is a tiny dot in the photo, cropping won’t make it any sharper but it will at least make it easier for the reviewer to find it within the image.
7 Likes
There is an open feature request to add cropping to web and iOS: https://forum.inaturalist.org/t/incorporate-basic-image-editing-tools-crop-rotate/1223. It has a pretty good amount of votes already, but more are welcome!
I expect that the new app version would include cropping since it is already available in the Android app.
1 Like
In the relevant feature request it notes
So this seems to say that it is being considered for the web, just low priority.
1 Like
In digital imaging terms, ‘artifacts’ are usually those little noisy clusters of pixels most noticeable in images (most JPGs, for example) along higher contrast edges. It’s an artifact of the compression process. Much more noticeable when you zoom in
Also, images that have been cropped or processed either in camera or out, to enhance detail or reveal detail in shadows or highlights, or contain large, subtle gradations, are also very prone to producing artifacting. Virtually all video frame captures will also include artifacts.
There are also many physical artifacts caused by lighting, reflections, lenses and even camera sensors.
Learning to control, avoid, and correct (when possible) artifacting is a definite skill to develop for photographers, even though the cameras are smarter than ever.
I suspect that the most common and frustrating type of artifacting for IDers is too few pixels, or too blurry or too noisy, or too low contrast details.
Pixel quantity and noise levels usually have some dependence on each other in most cases. Notice that I’m avoiding the term ‘resolution’ here as it’s a source of a lot of the confusion (and always has been!).
I would be willing to put together some basic guide (tip sheet with examples, perhaps) for iNat to help beginners learn how to deal better with this stuff.
If you’re an IDer and have some examples of both good and bad submissions on this topic, I would appreciate hearing from you.
2 Likes
just a reminder, there is no requirement to crop. I don’t reprocess my photos on the computer, though i do use the iphone zoom to ‘crop’. Please don’t harass people on inat to crop photos. Cropping photos also has other disadvantages. You gain nothing in terms of resolution and you lose the habitat context.
I do think it would be really neat if iNat had an internal crop or saved zoom type feature.
3 Likes
I agree, those shots are often crucial for ID, there’s many cases where the flowers look completely identical between species (and even across genera) and you need things like leaves or branches to get to species. Sometimes habitat also is the deciding factor with plants, e.g. if it’s clearly growing on top of sphagnum moss that could rule out one species in favor of another.
I think the whole cropping discussion is more about birds and insects and less relevant to plants, just because the structure of many plants means if you crop to the inflorescence you at the same time crop away other important parts like leaves or stems… so sometimes for plants the advice might be the exact opposite, don’t crop to just the inflorescence which seems to be the instinct of some observers.
5 Likes
Along these lines, is there a way to choose the order of the photos in windows version of iNat (I see it’s available on the iPhone app) without needing to finish submission and re-enter in edit mode? At least 80% of the time, iNat seems to select one of the supporting images to display as the primary. Am I missing something obvious???
1 Like
I liked their comment because to me a zoom in of pixels is less helpful to make out shapes. Its like the Hollywood forensic teams “enhance enhance!” You cant make pixels out what isn’t there. Ive seen plenty of cropping that is completely unhelpful and makes it harder to tell what is going on. However i do agree that on iNat, typically, cropping is helpful and the intent of this thread is to encourage people to drop extra stuff that makes IDing harder. Cropping still needs done with care though or it can easily be a mess of pixels. And for some organisms, it is really unhelpful (but this thread started about birds I think, where people aren’t going to crop out needed details)
1 Like
You can crop in the Android app.
I use GIMP to crop my photos. However, I find the metadata is preserved, as I simply use Select, Crop To Image Selection, and then export it as a new file.
1 Like
EDIT: Aargh, meant to be a reply to @gcwarbler https://forum.inaturalist.org/t/a-plea-to-crop-your-photos/35251/39?u=russellclarke
This is a much better explanation than I could write. Thanks, @broacher !
https://forum.inaturalist.org/t/a-plea-to-crop-your-photos/35251/44
Getting software to enlarge an area of interest (which is essentially what cropping achieves) without the downsides is not always straight forward, and therefore your mileage may vary. In many cases any benefit of a bigger subject is diminished lost by the poor quality of the result.
1 Like
I still think we seem to be talking past one another here, with significantly different concepts of what I (and others) intend for cropping advice. Just to illustrate my concept, I am going to point to one of my recent moth observations in AZ which I cropped significantly, rotated slightly, and lightened (in iPhoto on a Mac) prior to uploading. You see the result in the first image of that observation. I just added the original unedited version of the picture as a 2nd image on that observation–analogous to some of the moth and bird images that we see regularly on iNat. In the absence of my simple steps, the original observation (2nd image), while potentially identifiable, would need to be downloaded and edited to give others the chance to begin to ID the moth. My steps don’t create a “poor quality” result, quite the opposite. Believe me, I have uploaded my fair share of those as well!
And to reiterate what most of my iNat friends are aware of: I use an inexpensive point-and-hope Canon camera and outdated Mac software!–maybe both of those are to my advantage, actually.
15 Likes
that depends on the device and the photograph system. In my experience, using iPhone camera, and uploading the observation to the iPhone iNat app strips the metadata. Backing up observations with Google Photos on iPhone, then uploading to iNat on an Android device ensures intact metadata.
1 Like
Cropping has advantages. It makes what you want identified clear to both humans and CV. It can show the target organism at a larger size on the screen. It may save iNaturalist a bit on computer storage space. So it’s good to crop.
Cropping takes time, so sometimes I don’t do it even when maybe I should.
Cropping can create problems. If a tiny dot of a bird is cropped so it will fill the screen, all we see are blotches of colors. If you produce this, please also post the full image or one with only a minor crop. It can be no worse to identify than a thoroughly pixelated image, and may be better. My personal standard is that I should avoid cropping the image down to much less than 1/4 the original size, but that probably depends on camera and editing software.
Another problem can be cropping out information needed for identification (e.g. cropping a plant down to a flower, omitting leaves). However, I think the bigger problem here is not photographing enough to begin with. We often need leaves or the whole plant or the habitat for identification – keep that in mind.
I edit in Microsoft’s Paint because it comes with Office at no additional cost, and it’s easy to use. (Yes, I can hear jaws dropping as you prepare to protest, “That’s like doing surgery with a shovel!” However, sometimes a shovel is good enough.) I don’t loose the original date.
Using “Identify” on the website, there are little buttons with “+” and “-” that make the image lighter or darker, if that’s a problem.
3 Likes
There was a record recently referenced on this forum which was a bald eagle at some distance taken on a cellphone. As I recall the image of the bird was two pixels, one black and one white. It made Research Grade. Of course that was a special case … I wouldn’t expect most such pics to be IDed.
3 Likes
This is really important for plants. Also, take some closeups of leaves, stems and bracts as these are important for plant ID.
4 Likes
It seems your technology setup is sufficient to avoid the issues. Per my heavily qualified response, some people don’t have the same skill/tech/luck, and having been asked to ID heavily enlarged observations that have lost all their definition, I know they are using iNat.
1 Like
Zooming in and taking a Screenshot will cause you to loose metadata 100% of the time. I think that is probably why a lot of people have the problem, especially on phones. I am pretty sure my friend does that sometimes. You need to crop and then save as JPEG for most programs to retain metadata.
The only situation where a cropped area should have poorer quality than its source image would be if the crop has been re-compressed into a lossy format like JPEG. Anyone using something like lightroom will be fine unless they go out of their way to make a very compressed output file. On mobile, all crops (and other edits) in iOS are lossless, so there would be no loss of detail. It’s possible that some Android devices handle it less well.
What I DO see on recent iPhone photos is that crops often look terrible/blurry when the default camera is used, but this is not so much an issue with the crop as with the post-processing used on the photo as a whole. It just isn’t apparent until you zoom in or crop it. This can be avoided by using a 3rd party camera app that doesn’t use the post-processing fortunately.
5 Likes
The resolution is affected by the way iNat processes image submissions.
What you say would be correct if not for the fact that when you submit a photo to iNat, the maximum long dimension for an image is 2048 pixels. Which means that if you submit something larger – say 4, 5, or 6,000 pixels long, the image gets downsampled. And it’s there where the details will get tossed.
Let’s say you submit an uncropped 4096 pixel wide original. Your subject, uncropped, takes up a length of say 1,000 pixels. And you look at it and say, sure, that has enough detail for a solid ID.
Now you click on the uploaded image and observe that the pixel length of your subject is now only 500 pixels long, and the ID details are gone or too weakly defined in pixels.
If instead, you cropped the original image around the subject so that the cropped area is now 2048 pixels long. You upload that and… now the subject is the same number of pixels (1,000) long as it was in your original,uncropped photo. No pixels are lost, the detail preserved. And THAT’S why cropping closer pays off for the ID work.
I hope that wasn’t too harassing, but it’s why it makes sense to crop to subject for the purposes of creating an image to submit as an observation. Yes, habitat context is often very important. In which case, add the full photo as your second shot for the observation. That original image then gets ‘downsampled’ to 2048 and the detail pixels get jettisoned. But, it will still provide habitat info, if it’s needed.
And maybe you have a super high end camera with a 10000 pixel sensor! Great. Then you may even be able to crop to additional ‘zoomed in’ crops of the subject’s keying features to provide the IDer quicker, more positive ID.
I was recently told that I’m over explaining things around here, so I hope nobody is offended.
13 Likes