217-Aculea, de Vlaamse club van bijen- en wespenliefhebbers, presenteert een aantal online opleidingen

Coming here a bit late, but note that if you want to get a request body from hub.toolforge.org, you can set format=json in the querystring, and get the Wikipedia article URL in the response body at .destination.url. Example: https://hub.toolforge.org/P3151:506262?lang=en&format=json

Moreover, in case there is no article in Wikipedia for the desired language, the URL will be one of another Wikimedia project in the requested language, if available. Example: https://hub.toolforge.org/P3151:506262?lang=en&format=json

If you would actually prefer to get a URL on a Wikipedia in another language (rather than a URL from another Wikimedia project), you can specify site=wikipedia. Example: https://hub.toolforge.org/P3151:506262?lang=en&site=wikipedia&format=json

That image cache might be filling up, but adding a button to clear it out in the Android would be kind of redundant b/c the operating system already provides a way to clear out an app’s cache. For those of you for whom the Android app is using a ton of space (2+ GB?!), does clearing the app’s cache using that method reduce the disk usage (be careful to clear the cache and not the “storage” or I think it will equivalent to signing out)? If that does drastically reduce the amount of space used, is there really a need to implement a “clear cache” button in the app? IMO, an in-app setting to limit the size of the cache might still be useful, but it’s not clear to me from this thread whether people are asking for / voting for a limit or a “clear cache” optio

https://www.noardeast-fryslan.nl/sites/default/files/2021-03/Landschapsbiografie%20Noardeast-Frysl%C3%A2n%20-%20Definitieve%20Rapportage__lowres.pdf

https://www.saniwijzer.nl/media/stowa_saniwijzer_v2/org/1/documents/lopende%20projecten/201601_afvalwaterproject_drielanden_eindrapport-8mb.pdf

https://www.saniwijzer.nl/media/stowa_saniwijzer_v2/org/1/documents/lopende%20projecten/201601_afvalwaterproject_drielanden_eindrapport-8mb.pdf

European Model Autumn 2021

Opgenomen soorten geldt:

1. de taxonomische boom kloppen
2. minimaal 10 waarnemingen bestaan met 1 of meer fotos welke handmatig gevalideerd zijn met een groen vinkje binnen Europa
3. de foto`s moeten minimaal 400*400 pixels zijn
4. het maximaal aantal fotos voor 1 soort is 25.000

https://forum.inaturalist.org/search?q=elastic

If enough photos are present at the genus level, but not enough photos for any of the descendent species, then the genus will be placed in the training set. We call this approach a “leaf model.” :fallen_leaf::robot:

More details are available in Ken-ichi’s TDWG keynote from last year:
https://www.youtube.com/watch?v=xfbabznYFV0 9

Off the top of my head, some things we’ve taken from previous iNaturalist and other vision challenges include label smoothing, the Xception architecture, changes to how we generate our data exports & minimum number of images for each class, etc. Sometimes however contest winners have done things that we probably won’t do, like training on validation/test data (seems risky for a production model but perhaps worth it to win a contest) or training on larger image sizes (our models already take a long time to train).
The GPU does the lion’s share of the work in computer vision training, and the GPU we used (an NVIDIA RTX 8000) cost more than the rest of the PC put together.

Happily, NVIDIA is a supporter of Cal Academy of Sciences and iNaturali

1. search through places in iNaturalist to make sure a similar place doesn’t already exist. you can search by name at https://www.inaturalist.org/places. alternatively, go to the Explore page, zoom into an area on the map that roughly coincides with your watershed boundaries, and click the Places of Interest button to see if there are any places that encompass or are nearby this area. if an existing place doesn’t already exist, go to the next step. (if the place already exists, skip to the last step.)
2. open QGIS (download here: https://www.qgis.org)
3. go to Layer > Add Layer > Add ArcGIS FeatureServer Layer…
4. in the Server Connections popup, click the New button
5. in the new connection popup, provide a name of your choosing. in the URL box, use https://njwebmap.state.nj.us/arcgis/rest/services/Features/Hydrography/MapServer. you don’t need to fill anything else in this case. click ok.
6. back at the Server Connections pop-up, select your new connection, and click the Connect button. this brings up a list of layers on the server. in this case you’ll select 17, and click add. that will add the layer to your project.
7. go to Layer > Open Attribute Table. in the new window, click the Select/Filter Features using form button (or ctrl+F). enter 02030105130 in HUC11 box (or Lawrence Brook in the W_NAME box), and then click Select Features. close the window.
8. back at the project window, your feature should now be selected. go to Edit > Copy Features. then click Edit > Paste Features as > New Vector Layer.
9. in the Save Vector Layers as box, select KML as your format. then click on the […] button next to the File Name field. Choose a path on your local machine, and provide a file name of your choosing. click OK.
10. you can close QGIS at this point.
11. you now have a KML that you can load into iNaturalist. go to the places page and click on the the Add a New Place button (available only to users with at least 50 verifiable observations, i believe).
12. in the Create a Place screen, provide a descriptive name, and select your KML file. add a parent if it exists, and select a place type, if you like. then save your place. you now have a new place.
13. now go to your project (or create one if you haven’t already), and use this new place in your project setup.

Off the top of my head, some things we’ve taken from previous iNaturalist and other vision challenges include label smoothing, the Xception architecture, changes to how we generate our data exports & minimum number of images for each class, etc. Sometimes however contest winners have done things that we probably won’t do, like training on validation/test data (seems risky for a production model but perhaps worth it to win a contest) or training on larger image sizes (our models already take a long time to train).
Yep, a single model. We’ve considered stacked and split models, but we haven’t seen evidence that they are any better than a single monolithic model, so we’ve stuck with the simple approach. Happy to be corrected if anyone’s seen or done research that points in another direction!

Yep, a single model. We’ve considered stacked and split models, but we haven’t seen evidence that they are any better than a single monolithic model, so we’ve stuck with the simple approach. Happy to be corrected if anyone’s seen or done research that points in another direction!
I haven’t experimented with transformers, but after their strong showing at the 2021 iNat Challenge (and many other vision challenges), I’ll certainly take a look.

https://www.youtube.com/watch?v=mIlPcQchz9A

• de taxonomische boom kloppen
• minimaal 10 waarnemingen bestaan met 1 of meer fotos welke handmatig gevalideerd zijn met een groen vinkje binnen Europa

• de foto`s moeten minimaal 400*400 pixels zijn
  het maximaal aantal fotos voor 1 soort is 25.000

https://www.vogelwarte.ch/assets/files/projekte/entwicklung/zustandsbericht%202021/Zustandsbericht%202021_en_low.pdf
The annual publication “The State of Birds in Switzerland” (in English, French, German and Italian) summarises the results of various monitoring projects, conducted with the support of more than 2,000 volunteers in all parts of the country. The 2021 report focuses on EBBA2 and its implications for Switzerland. It also takes a closer look at trends in scarce breeding birds like Western Yellow Wagtail, rare passage migrants such as Glossy Ibis, and local winter visitors such as Whooper Swan. You can also explore interactive graphics from the Swiss Bird Index SBI® and the breeding bird index for each species until 2020