Real-time Computer Vision predictions in Seek by iNaturalist 2.0

I think the latest version of Seek has iNat upload capability, due to high demand. I think it’s not ideal though, because there is no GPS location picked up by Seek app, so you have to place location pins manually. Search "Seek: in the forum to find the topic about the change to Seek to allow iNat uploads… it describes how to set it up as well

1 Like

@tiwane Any chance in one of the next updates a focus function will be implemented in the camera mode? I continuously fail to capture insects sitting on a plant, because the focus in on the background.

2 Likes

We are working on both pinch-to-zoom and tap-to-focus. Looking pretty good in beta.

8 Likes

some weeks ago I saw that tap-to-focus is already implemented in the iPhone app. Will there be an Android update as well soon?

We’re working on it. Apparently it’s much tougher to implement on Android.

3 Likes

Just a quick feedback: I see that a new Android update (Edit: Seek version 2.3.3 on Android 9 Pie) has been released, including the option to flag wrong IDs and a ‘tap-to-focus’ feature. However, I could not see an improvement in getting stuff into focus, all it does is getting blurry upon tapping and re-focussing on the background (tested it in different environments).

For the flagging - it would be great to label wrong IDs which are already in my Poké Ball …- ahem… I mean stored under my observations, i.e. after ‘collecting’ them. In the field, a substantial purpose of using Seek is to have unknown objects identified. How is then one to decide whether the ID is trustworthy?

Lastly, I saw that for observations w/o vernacular names, the scientific name is converted and capitalized (that false-IDed Chinese moth is still in my observations…:roll_eyes:). That wasn’t the case in the previous version (see here: https://forum.inaturalist.org/t/improving-translation-in-seek/4758/3?u=carnifex)

1 Like

Thank you for the feedback. Probably best to convert the first and third issues into bug reports so they don’t get lost here. The second issue is something we’re working on. Please include which version of Seek you’re using.

A review about some nature apps, e.g Seek. https://www.rtlnieuws.nl/lifestyle/artikel/4826616/vogels-spotten-planten-herkennen-apps-seek-inaturalist-hortus-ibirds

Surprised, cause I should expect more attention for iNaturalist app.

1 Like

i only seem to get ids when i have access to wifi. is it possible to use seek in airplane mode? i am trying to use it with students who have school provided ipads during covid distance learning. i want to encourage them to get out for a walk and to collect identifications, but seek doesnt seem to save when i get away from wifi

1 Like

No, you need wifi or cell data connection - Seek has to access it’s ginormous database of photo patterns to match. However, you can take a photo with the tablet, then feed that to Seek when you are back in wifi range. Erroneous answer withdrawn (see below).

Did you know?
• No Internet is required to
identify and add species to your
observations!

This identification does not leave
your device or require an internet
connection to work

Accoording to the documentation http://static.inaturalist.org/wiki_page_attachments/SeekUserGuide2020.pdf no internet is required for Seek.

Sorry, my mistake. I’ll delete my previous comment.

The model Seek uses is simpliefied to fit on a mobile phone so it works not as good as the other apps that access the complete online model.

@ekohner It should work without wifi.

1 Like

https://www.inaturalist.org/posts/25510-vision-model-updates#comments

Citaat

Smaller

The mobile version of the model needs to be small in terms of file size to minimize the amount of data app users would need to download. Smaller models can also be used by more devices as they need fewer resources to run (e.g. memory, battery), and can generate results faster, which is important for Seek’s real-time camera vision results. These models take a lot of time and money to train, so we also wanted a model that could be simultaneously trained to produce a large web-based version and a smaller version for use in mobile devices.

Unfortunately, shrinking the file size like this slightly decreased model accuracy compared to the larger web-based version (kind of similar to image compression), and we found that was an unavoidable tradeoff. We take this into account when processing the model results, and on average for a similar error rate, the mobile version might recommend a taxon at a higher taxonomic rank than the web-based version. The taxon results we show to users shouldn’t be less accurate, but they may be less specific.

1 Like

Is there anything in the works to be able to upload this to a digital camera or other device that is not a phone? The reasoning behind this is that having a device for iNaturalist would allow classrooms to purchase them for outings and allow people to use them at nature centers and it would hopefully be less expensive than buying phones. Also, kids would get into less mischief with something that just took pictures and identified species.

1 Like

Hi, welcome to the iNat Forum!

It’s an interesting idea, though there are a lot of inexpensive phones and tablets that are much cheaper than most digital cameras.

3 Likes