I’m struggling with a recurring performance issue when uploading large batches of observations.
My setup: Mac user, using an OM System TG-7. I pre-resize all my photos to 800 KB to make the upload easier.
The issue: When I upload a batch (e.g., 100 photos), the first 40 suggestions are near-instant. However, once I pass the 50-observation mark, the AI suggestion tool starts lagging significantly. It can take minutes to load a single identification suggestion, making the whole process take over an hour for a typical field day.
What I’ve tried:
Switching browsers (Chrome to Opera).
Reducing batch sizes (uploading 80 photos at a time instead of 150).
Resizing photos beforehand.
Despite this, the lag always returns. It feels like a memory leak or a browser cache issue. Has anyone else experienced this? Are there any workarounds or better workflows for power users handling hundreds of observations daily?
For my workflow, that would be a huge upload attempt. I normally do bulk uploads in batches on the order of 10 to 20 observations at a time, maybe max of about 25. There is sometimes a few seconds delay in CV’s suggestions showing up, but never on the order of minutes. Using Chrome with MacOS Sequoia 15.7.3.
The only solution I’ve found is uploading about 50 each time and letting it rest for a few minutes between uploads. I think it’s related to the API limits for the model or something like that. Because even in the same upload if you wait it works again
I’ve experienced the same issue! Glad to hear it’s not just on my end. Since I’m stuck in premed Anki hell, what I do is upload batches of 50-100 photos in between doing 100 Anki cards.
I looked at a small random sample of your images uploaded in 2025-26 and all of the images, at full size, are in the 1.2 to 2 mB range, not 800 kB. So I don’t know that your resizing is accomplishing what you desire. Nor do I think images as small as 800 kB are necessary. Your images of butterflies and sealife are gorgeous!
Same here. One “trick” that sometimes works is just submit when it slows down. You will get a warning about not all observations having identificaitons. Cancel the submit, then start to go back through the AI suggestions. Sometimes it works, sometimes it doesn’t. I’ve also found that my 3rd or 4th batch of 50 will hit this issue sooner. May have to move to 20-30 at a time.
It happens all the time with larger batches of observations. It is not, as far as I can tell, due to image sizes. Instead you are possibly getting rate-limited on some API calls. Often if you just wait a bit will improve.
If you are somewhat tech savvy you can check your browser’s console (network tab) to see what is happening. You will likely see some calls to various iNat API endpoints returning some form of 4xx errors.
I use Firefox, oftentimes on an extremely slow network. There are a few workarounds you can try:
When you do a batch of more than 40, it takes a significant while for all the meta-data to upload as well. Only start using the suggestion tool once all of the images have had their meta-data uploaded.
Work in batches of 30 observations max at a go, open multiple windows so all the meta-data can load while you are busy with the first batch.
I’ve never seen that picture size has an effect on this, I honestly don’t think it is required to lower the quality.
Try uploading when you have a really stable internet connection, right now I’m on 2.4Ghz signal and an upload of 60 images will take around an hour, on a fibre network that reduces to about 10 minutes
As others have mentioned, yes known issue. Exasperated by slow upload speeds. If you click the species name box, then click off, repeat with multiple observations, the system seems to work in the background, and the next time you click on that observation some time later it will pop up immediately. I use this to “queue up” the cv when uploading with a slow connection, while working on other parts of the obs such as notes or fixing locations.
this is probably intentional. I often upload one-night’s-moths in one go, which can be 100-300 at a time, particularly in good Summer nights or tropical locations. How I do it?
(1) upload all the photos and do nothing until the metadata has loaded for all, then add location, tags etc. and use AI identification as far as it’s fast. When it stops, go drink a coffee and come back in 10min. Continue until it’s stopping again. Keep repeating. Just take it easy. This is my choice if I am at home.
(2) upload all the photos and do nothing until the metadata has loaded for all, then add location, tags etc. and use AI identification as far as it’s fast. When it stops, upload everything without ID. Ignore the warning. When you go back to all your “unknowns”, you will see that you can use AI identification for hundreds of observations in one go! Disadvantage: usually some busybody will interfere and be faster than you with ID on your own uplaods (which I find kind of annoying). This is my choice if I am afraid the internet connection or electric power could interrupt (remote locations, mountains, jungle, desert).
but not (3): splitting it up into smaller batches I find less convenient, because I often have to combine photos of the same species. These photos are often taken hours apart, e.g., at the beginning of the night + towards dawn. It also is more work, because location, tags and remarks have to be added for each batch separately. In addition, I make more mistakes when choosing the photos: I “correct” the date+time after midnight usually back to the previous date (without time). I find this necessary because I think the entire night should be “one entity”. Therefore I prefer methods (1) or (2).