While importing some images from iNaturalist to Wikimedia Commons, I noticed that most of the images have been downsampled to 2048 pixels wide (as the “original” size). While I’m sure that iNaturalist doesn’t want to store hundreds of thousands of 20 megapixel photos, 2048 is pretty small, and it’s a shame that we’re losing the original higher resolution versions, especially when they are freely licensed and could be used elsewhere. Even a modest increase to 3000 would make a significant difference when reusing images in print, for example.
While the high res images being available is a great thing…but they sure do hurt people that are rural and/or have limited data allowance to use each month and no options for any other internet service. It really restricts how much you can view and enjoy iNat when the large images loading use up precious data allowance.
9 Likes
I understand the cap is partly to avoid iNat becoming an image hosting service, with the idea that people can link to higher-res versions (e.g. on flickr). I guess there are tradeoffs? A related policy is that iNat supports animated gifs but not video – and encourages links to youtube/vimeo/etc where it could be helpful.
7 Likes
Yes! Rural here with no cable and terribly slow satellite internet available. I’m now doing unlimited data on my phone and using a hotspot but that is restricted down to 600kb/s. I get so frustrated by huge files especially when it is huge because they didn’t crop so I can’t ID from the size that shows up on the observation and have to view full size. Then I’m waiting for ages for the top half of the photo that is just vegetation to load so I can see the subject of the observation. 
12 Likes
I’d love to have higher-res photos on iNat, but IMO the infrastructure and data connection issues supercede the image quality one.
Also, @zygy don’t forget to vote for your own request. :-)
9 Likes
How should this work?
- You upload a picture to iNaturalist.
- iNaturalist checks if they are freely licensed and if so iNaturalist stores the original file in Wikimedia Commons.
- An observation field is updated to store the relation with the original file in Wikimedia Commons.
- iNaturalist resizes the original file to 2048 pixels wide and stores it in its own database.
- If the observation becomes Research Grade the tags of the file in point 3 are updated with the taxon name.
7 Likes
Yes. Please stop compressing images from Trusted users.
So much quality images are being unnecessarily ruined. Images from trusted users should not be compressed.
1 Like
To be clear, I’m not suggesting displaying full resolution images on the observation pages. I’m suggesting that a higher resolution version be retained and made available specifically for people who want to download the higher resolution version from the photo page or API.
3 Likes
A lot of my time on iNaturalist is spent zooming in on photos in a separate browser window to try to make out barely visible or entirely invisible features. It was a bit disappointing to realize that iNaturalist is deliberately deleting a lot of the information that I’m squinting to try to make out.
Of course, I understand that server space and bandwidth aren’t free. There may be better solutions, though. For instance, if an image is already on flickr… why not use the version hosted there? When I click on an image to make it bigger, just show me the full resolution version that’s already hosted. Yes, I can get to it from iNaturalist as-is, but it’s a lot of clicks and page loads for each image. And if I’m going through observations rom other users, it’s a lot of clicks just to tell if there are larger versions.
Or maybe make access to full resolution images a paid feature. I’m sure some people would be irritated, but if the cost of server space and bandwidth is the limiting factor, get people to pay for it. I would. And since no one can download full resolution images now… who would be harmed?
With regard to download times for people with slower internet connections, a user setting along the lines of “display maximum images at … x … pixels” would probably be a good solution. Making data access faster by deleting a lot of the data is not a good solution.
Since bandwidth and hosting costs are now free (https://www.inaturalist.org/blog/49564-inaturalist-licensed-observation-images-in-the-amazon-open-data-sponsorship-program/) for iNat photos under CC licenses, those CC photos could be stored at higher resolution with no additional cost to iNat except for development cost of enabling different sized uploads. The other limitation, data use for mobile users, could be addressed with a switch in the app settings for low-resolution uploading, although those mobile users might still end up downloading higher resolution photos while looking through photos of a species, for instance.
7 Likes
I’d still prefer iNaturalist to encourage cropping photos. I’ve seen too many large photos with the actually observed animal/plant just being a tiny spot somewhere in the photo (not necessarily the center).
And just for the purpose of demonstration, I uploaded the full 6,000 x 4,0000 pixels photo of a lady bug taken from 1 meter distance, which was then resized by iNat to 2,048 x what ever pixels. See observation
https://www.inaturalist.org/observations/131767933
The 3rd photo is the big one, the second photo is the cropped version of the third photo.
While it is true that even that big (and re-sized) photo is good enough to identify the species, you can clearly see the waste of ressources. Also the ressources of the identifiers might be wasted, when they have to search for the bug (which is easily visible in my example).
6 Likes
Even with cropping, I would really like larger image sizes. I crop as much as possible every iNaturalist photo I take and most of the crops are still much larger than 2000px. And I only use a phone camera.
4 Likes
I agree, I’d love a basic crop tool incorporated into the web uploader. For many photos, I don’t need high quality, just trimming away the excess in 4 seconds to help identifiers with a closer view. This would meet some (but not all) of the same goals as increasing the max image size for many users.
I do think if max size were increased, that some sort of toggle to allow downloading lower resolution would be needed so as not to burden those with low internet speeds/poorer quality access.
3 Likes
iNat has said on an earlier thread - won’t be offering cropping. We must use existing photo tools.
1 Like
Regardless if the iNat app as such comes integrated with a cropping tool or not, my concern is more about the FAQ / tutorials, e.g.:
https://www.inaturalist.org/pages/responses#crop
says:
Would recommend only using this with new users and only if the organism is actually unidentifiable. Users are not required to provide cropped photos.
While I agree with cropping not being required, I would prefer to see it more “encouraged”, advertised as a “good practise” or however you may call it.
6 Likes
On the other hand, there are a couple of photos with exceptional quality. I understand that someone might want to store them at full resolution for purposes beyond identification (like printing as mentioned in the first post). When is a photo “exceptional”? I do not want to define it, let’s try to have reasonable users decide themselves.
So, there could be a check mark visible only when a large photo is about to be uploaded (and license requirements are met), asking if the user deems it “exceptional” and the photo should be stored at full resolution.
Not checked by default, because then almost every low quality photo would get too much storage space.
That could be worth a try.
Keep in mind that it is not resolution per se that we care about. It is the trade-off between image quality, and the resulting file-size that matters. And resolution is only a part of the image quality choice, along with the compression algorithm and particular settings used. In general I would say it is better to increase the JPG compression rather than decreasing the resolution, but ideally this should be on a sliding scale such that the higher resolution, the more compression is applied (i.e. the lower the JPG quality setting is set). For smaller cropped images, you want less compression applied where every pixel may count. I would say iNaturalist does a better job of getting the balance right than many other web services do, though there is always room for improvement.
Apart from using a sliding-scale compression setting, other things they could look at would be whether they’re using a modern compression algorithm like MozJPG already and fine tuning the parameters within that to be best suited to wildlife images. Beyond that, the next big gain I think could come from implementing a more modern format. JPG XL looks like the best option here since it can be freely converted back to JPG again without any quality loss. It is already supported by most desktop browsers (but needs to be enabled), but not mobile browsers so much yet I believe, so perhaps one to consider for the future (or perhaps only implemented for the largest size “original” image).
Squoosh and Caesium are both good open source image compression tools for reference:
https://github.com/GoogleChromeLabs/squoosh/
https://github.com/Lymphatus/caesium-image-compressor
JPG XL is currently in beta on Squoosh.
As others have mentioned, the far bigger issue I see in practice is with mobile app and less technical users not cropping images prior to uploading. So I strongly agree with those suggestions around encouraging users to do this.
4 Likes
It seems like iNat has at least these groups of users with respect to photo quality:
- High quality enthusiasts - more resolution, higher quality, photoshop users. Maybe folks in topic: Data Hoarders Anonymous?
- Medium quality - smart phone or point-and-shoot camera users, maybe not aware of advanced photography terms, technology, etc. Quality is driven by equipment defaults.
- Identification quality - people who crop photos with just enough detail for identification. Mix of people.
- Low quality - blurry photos from far away that get pixelated when zoomed in. New users who have very high expectations about image enhancement after watching crime drama shows like CSI or NCIS, probably?
I’m in the identification & medium quality group, so my bias is towards lower resolution cropped photos since they load faster and don’t need zoom-in to see diagnostic features. It’s still nice to see high quality photos too, where a photographer took time and consideration to frame the subject.
3 Likes
4: All of the above. I use a gradient of quality standards and gear in my never-ending quest to reduce my “dang, it got away” list.
4 Likes
Good point! I forgot about the safety shot, it’s thankfully something that’s not needed with plant photos.