18. Beeldherkenning bij iNaturalist en Wildcamera's Citizen Science Snapshot

iNaturalist gebruikt beeldherkenning en bracht elk kwartaal een nieuwe versie uit. Tegenwoordig kost het doorrekenen van een model zoveel tijd en gebruiken ze zoveel foto's dat het nu eens per jaar gaat worden denk ik. Wel apart dat men gelijk denkt dat vrijwilligers overbodig zijn.

https://www.zooniverse.org/projects/y-dot-liefting/snapshot-hoge-veluwe?language=nl
https://www.nrc.nl/nieuws/2019/10/25/gesnapt-intelligente-cameras-leren-welke-dieren-voorbijlopen-a3978047
https://www.zooniverse.org/projects/y-dot-liefting/snapshot-hoge-veluwe?language=nl
https://diopsis.eu/resultaten/
https://www.nu.nl/nucheckt/5406416/nucheckt-onwaarschijnlijk-kledingkleur-invloed-heeft-wespen.html

Wildonderzoek samen met publiek

Om overlast en schade door wild te voorkomen, beschikt het Nationale Park de Hoge Veluwe over een uniek wildmonitoringssysteem. Een netwerk van cameravallen registreert continu de activiteit van wilde dieren, waarbij Wageningen Universiteit de gegevens gebruikt voor onderzoek. Het netwerk levert miljoenen foto’s op, waarvan slechts een deel verwerkt kan worden. Met dit project wordt het huidige systeem aangepast om citizen science in te zetten voor de verwerking van de foto’s. Om mensen te trainen komt er daarnaast een online leermodule en game. Het project levert niet alleen essentiële informatie op voor het natuurbeheer, maar leidt ook tot educatie en grotere betrokkenheid bij de natuur.

Project: Snapshot Hoge Veluwe - Wildonderzoek met publiek, Stichting Het Nationale Park De Hoge Veluwe & Wageningen Universiteit

Ruseler: „ In 2013 besloten we de omrastering van de Hoge Veluwe te verlagen, en we wilden weten wat het effect op de verspreiding van de soorten was. Met Wageningen Universiteit plaatsen we 55 camera’s, inmiddels hebben we het opgeschaald naar een project met 70 camera’s, om jaarrond het gebied te kunnen monitoren: waar bevinden zich welke soorten, wat doen ze, hoe verhouden ze zich tot elkaar?

„Zo bestaan er in het buitenland al projecten als ‘Galaxy Zoo’ en ‘Chimp & See’, om meer te weten te komen over het heelal en mensapen, en van Snapshot bestaat er ook een Serengeti-versie.

In de winter hebben de dieren een dikkere vacht en werkt de infra rood camera ook minder goed omdat de dieren pas veel later gezien worden. In de zomer worden er wel vijf keer zo veel fotos gemaakt door de 77 wild cameras.
De maximumafstand waarover een camera een dier kan waarnemen is ongeveer 25 meter. Jansen: „Natuurlijk kun je op de weides verder kijken dan in het bos, omdat er lagere begroeiing is. Maar ze werken op elk type terrein naar behoren.”

„Zoogdieren zijn geruislozer en moeilijker waar te nemen dan vogels”, vult computerwetenschapper Benjamin Risse van de Universiteit van Münster aan. „Vaak moeten zoölogen het doen met pootafdrukken en de incidentele roadkill. Daarom is machine learning zo interessant.’

Bij zulke machine learning wordt het computeralgoritme ‘getraind’ in het herkennen van beelden. Netwerken die daarin gespecialiseerd zijn worden CNN’s genoemd: convolutional neural networks. Risse en zijn collega’s hebben ook het CNN voor Snapshot Veluwe ontwikkeld. „Eerst hebben we dag- en nachtafbeeldingen van negen soorten laten zien: wild zwijn, das, vos, haas, konijn, schaap, damhert, ree en edelhert. Daarna hebben we eerst handmatig aangegeven waar op de foto de dieren te zien waren, zodat de computer dat kon leren.” Dat het niet altijd goed gaat, bewijzen de false positives: foto’s waarop de camera een dier waarneemt, terwijl er helemaal niets te zien is. „Daarom zijn dubbelchecks door mensen altijd nog belangrijk: mensen zijn vooralsnog nog de beste patroonherkenningsmachines.” Ook is het belangrijk om gokwerk door een CNN te voorkomen. „Je wilt niet dat je een bestand hebt met 22.727 zwijnenfoto’s en 292 dassenfoto’s en dat de computer daarom maar op een zwijn gokt.” Toch zijn er al soorten waarin computers uitermate goed getraind zijn, zegt Risse. „In het herkennen van hondenrassen zijn CNN’s vaak al beter dan mensen.”

Vroege vogels schreef: “Sinds een jaar staan er verspreid over Nederland ‘slimme’ camera’s die de stand van de insecten meet. Het project is gestart naar aanleiding van een Duits onderzoek waaruit bleek dat in 27 jaar tijd de insectenstand, gemeten in biomassa, met 75% is achteruit gegaan. Om de vinger nu aan de pols te houden is dit systeem, Diopsis, geïntroduceerd. De eerste resultaten zijn per camera gepubliceerd op de website. De analyse van de gegevens moet nog plaatsvinden. Naar verwachting komen volgend jaar ook intelligente insectencamera’s op de markt die iedereen in z’n tuin kan plaatsen. Dan kan je behalve meedoen aan het onderzoek ook zelf te weten komen wat er dag en nacht in je tuin aan insecten voorkomt.” samenwerkingsproject tussen EIS, Naturalis, Radboud Universiteit en Cosmonio, dat twee jaar geleden begon.

De camera is waterdicht en heeft geen ventilatie. „In de zomer werd het soms bloedheet: 80 graden Celsius, terwijl de chip er al mee ophoudt bij 85 graden Celsius. Dankzij die zonwering werd het nog maar 70 graden.” Wel wordt elke foto met die van 10 seconden eerder vergeleken. „Als er geen verschil tussen zit, gooit de computer de dubbele weg. Dat scheelt een hoop data, zeker omdat nachtvlinders vaak wel uren kunnen blijven zitten.”

https://www.inaturalist.org/journal/ahospers/18629-computer-vision-artifical-knowledge-links

https://www.zooniverse.org/projects/y-dot-liefting/snapshot-hoge-veluwe?language=nl
https://www.nrc.nl/nieuws/2019/10/25/gesnapt-intelligente-cameras-leren-welke-dieren-voorbijlopen-a3978047
https://www.zooniverse.org/projects/y-dot-liefting/snapshot-hoge-veluwe?language=nl
https://diopsis.eu/resultaten/
https://www.nu.nl/nucheckt/5406416/nucheckt-onwaarschijnlijk-kledingkleur-invloed-heeft-wespen.html

18C Computer Vision Artifical Knowledge Links

24. Herkenning van Soorten met Model 5 (Voorjaar 2020) in iNaturalist (TensorFlow 2)
Waarneming.nl

  1. December 2017 Photos van Voor 2017
  2. December 2019 Photos van Voor 2018
  3. December 2020 Photos van Voor 2019

    Globaal waren de oude versies:

    1. May 2017 Model 1 2-20 photos per species
    2. Aug 2017 Model 2 40 photos per species
    3. Jan 2018 Model 3 40 photographers per species
    4. Feb 2019 Model 4
    5. Sep 2019 Model 5 <1000 photos per species/li>
    6. Mar-Jul2020 Model 6, Tensor Flow2 25,000 taxa
    7. Mar-Jul 2021 Model 7, Tensor Flow2, 25,000 taxa +25.000 leaves , 21milj training photos , 1 NVIDIA RTX 8000 GPU, with hybrids
    8. Oct-Jul2022 Model 8, Tensor Flow2, 47.00 taxa +25.000 leaves, 25milj training photos , 3NVIDIA RTX 8000 GPU, 4 times Memory, Label smoothing, without hybrids

    Referentiewaarnemingen er gebruikt worden (5000 of 40)
    https://www.inaturalist.org/posts/59122-new-vision-model-training-started
    https://groups.google.com/forum/#!topic/inaturalist/K9nJOC0Cjss
    Referentiewaarnemingen er gebruikt worden (5000 of 40)

    1. https://groups.google.com/forum/#!topic/inaturalist/K9nJOC0Cjss
    2. https://www.inaturalist.org/pages/help#cv-taxa
    3. https://www.inaturalist.org/pages/help#computer-vision
    4. https://www.inaturalist.org/pages/help#cv-select
    5. https://www.inaturalist.org/blog/31806-a-new-vision-model
    6. FWIW, there's also discussion and some additional charts at
      https://forum.inaturalist.org/t/psst-new-vision-model-released/10854/11

    7. https://www.inaturalist.org/pages/identification_quality_experiment
    8. https://www.inaturalist.org/journal/loarie/10016-identification-quality-experiment-update
    9. https://www.inaturalist.org/journal/loarie/9260-identification-quality-experiment-update
    10. about a rare species, but the system might still recommend one based on nearby observation
      https://forum.inaturalist.org/t/identification-quality-on-inaturalist/7507

    11. https://github.com/kueda/inaturalist-identification-quality-experiment/blob/master/identification-quality-experiment.ipynb
    12. "nearby" means near in space and time
      The model became more efficient in sedges and grasse
      the vision model does not itself incorporate non-image data other than taxon IDs
      b/c because
      https://www.inaturalist.org/blog/25510-vision-model-updates ("taxon and region comparisons" 20190614)

    13. https://distill.pub/2020/circuits/zoom-in/ ("connections between neurons")
    14. https://www.inaturalist.org/projects/flora-of-russia/journal/31726
    15. https://www.inaturalist.org/posts/31726-
    16. https://forum.inaturalist.org/t/provide-relevant-geographic-data-confidence-level-accuracy-scores-with-ai-suggestions/9226/2
    17. https://forum.inaturalist.org/t/range-covered-by-the-seen-nearby-feature/2849/5
    18. --
      https://forum.inaturalist.org/t/use-computer-vision-to-annotate-observations/3331

    19. https://forum.inaturalist.org/t/what-image-s-are-used-for-training-computer-vision/3307/6
    20. Interessant artikel over de Artificial Intelligence achter Inaturalist: https://news.mongabay.com/wildtech/2017/08/smartphone-app-combines-computer-vision-and-crowdsourcing-to-explore-earths-biodiversity-one-photo-at-a-time/
    21. = = = Nov2020
      https://forum.inaturalist.org/t/better-use-of-location-in-computer-vision-suggestions/915/32

    22. https://github.com/inaturalist/iNaturalistAPI/blob/main/lib/controllers/v1/computervision_controller.js#L368
    23. Common ancestor for the top 3 raw results

      https://github.com/inaturalist/iNaturalistAPI/blob/main/lib/controllers/v1/computervision_controller.js#L368

    24. https://www.inaturalist.org/journal/ahospers/28858-funding-infrastructure-costs-images-on-amazon

      m looking for a way of finding observations without coordinates. Many of these have Location Notes, so it is basically lacking Longitude or Latitude that I am looking for.
      I am not interested in those with Latitude = 0 or Longitude = 0 (see https://www.inaturalist.org/projects/null (which is very inappropriately named, as I am looking for NULLS but this project identifies zeros instead - nulls have no data (value unassigned, or empty, or missing), but 0 is a specific datum - zero - like any other value - and not a “null”)).
      At present for this user, filtering on verifiable=false gives me more or less what I want, but conflates these with any Data Quality criteria, not just missing coordinates.
      https://www.inaturalist.org/observations?place_id=any&subview=grid&user_id=ahospers&verifiable=false 1

      I added a very basic search to atlases in response to Jane’s feature request https://www.inaturalist.org/atlases 6. So now if you wanted to see all ‘marked’, ‘active’ atlases of taxa in the LIliaceae you’d do https://www.inaturalist.org/atlases?utf8=✓&filters[taxon_name]=Lilies&filters[taxon_id]=47328&filters[is_active]=True&filters[is_marked]=True

      The out-of-range is vestigal, we don’t display it anywhere anymore (except the old filter menu thats still on https://www.inaturalist.org/observations/loarie 1). It worked directly on the taxon-range, rather than using atlases



    25. https://forum.inaturalist.org/t/identification-quality-on-inaturalist/7507


    26. AI Model 7 . July 2021

      The number of taxa included in the model went from almost 25,000 to over 38,000. That’s an increase of 13,000 taxa compared to the last model, which, to put in perspective, is more than the total number of bird species worldwide. The number of training photos increased from 12 million to nearly 21 million.
      Accuracy
      Accuracy outside of North America has improved noticeably in this model. We suspect this is largely due to the nearly doubling of the data driving this model in addition to recent international growth in the iNaturalist community. We’re continuing to work on developing a better framework for evaluating changes in model accuracy, especially given tradeoffs among global and regional accuracy and accuracy for specific groups of taxa.

      The recent changes removing non-nearby taxa from suggestions by default have helped reduce this global-regional accuracy tradeoff, but there’s still more work to do to improve how computer vision predictions are incorporating geographic information.
      https://www.inaturalist.org/blog/54236-new-computer-vision-model
      Participate in the annual iNaturalist challenges: Our collaborators Grant Van Horn and Oisin Mac Aodha continue to run machine learning challenges with iNaturalist data as part of the annual Computer Vision and Pattern Recognition conference. By participating you can help us all learn new techniques for improving these models.

      Start building your own model with the iNaturalist data now: If you can’t wait for the next CVPR conference, thanks to the Amazon Open Data Program you can start downloading iNaturalist data to train your own models now. Please share with us what you’ve learned by contributing to iNaturalist on Github.


    27. BIODIV Next
      Een presentatie hoe een be-nl model samengesteld is https://observation.org/download/Biodiv%20Next%20-%20Dutch_Belgian%20species%20ID%20.pptx
      Hierarchisch Model Ensemble is nauwkeuriger dan een singe model, mogelijk omdat bij 16.000 soorten te veel keuzes gemaakt moeten worden (Inception-v1, Inception-v3,Inception-v4, ResNet-18,, ResNet-34 , ResNet-101, GoogleLeNet, BN-NIN, GG-10)
      Performance vs Voorkomen


    28. https://www.inaturalist.org/blog/archives/2022/05
      https://www.inaturalist.org/blog/66531-we-ve-passed-100-000-000-verifiable-observations-on-inaturalist
      https://www.inaturalist.org/blog/63931-the-latest-computer-vision-model-updates

    29. AI Model 8 . May 2022

      In 2017 the amount of recognised species was 20.000 and now it is still.....20.000?
      https://www.inaturalist.org/pages/help#cv-taxa
      FWIW, there's also discussion and some additional charts at https://forum.inaturalist.org/t/psst-new-vision-model-released/10854/11
      https://forum.inaturalist.org/t/identification-quality-on-inaturalist/7507
      AI Model 4 . July 2019 included 16,000 taxa and 12 million training photos.
      AI Model 5 . July 2020 included 25,000 taxa and xx million training photos.
      AI Model 7 . July 2021 included 38,000 taxa and 21 million training photos. Training job in October 2021, we planned to train a AI Model 7 . May 2022 on 47,000 taxa and 25 million training images but finished with er 55,000 taxa and over 27 million training images.
      March 2020
      https://www.inaturalist.org/blog/31806-a-new-vision-model
      Juli 2021
      https://www.inaturalist.org/posts/54236-new-computer-vision-model
      2022
      https://www.inaturalist.org/blog/63931-the-latest-computer-vision-model-updates





    30. Posted on 21 de outubro de 2020, 11:37 AM by ahospers ahospers

      Comentários

      Google started with FreeBase

      Language Support
      Language support in EOL v3 is in continuous development, but many features are internationalized. Here's where things stand at the moment:

      The interface- navigating EOL in different languages:

      Thanks to our collaborators at translatewiki and their corps of volunteer translators, the full EOL basic interface navigation is available in Arabic, Brazilian Portuguese, English, Finnish, French, Macedonian, Piedmontese, Traditional Chinese and Turkish. Read more about becoming a volunteer translator.

      Common or vernacular names for taxa:

      We have harvested the common names holdings of the wikidata, which include just over 291,000 names in 279 languages. We also have >93,000 common names in 130 languages added by valiant EOL members to fill gaps they observed over the past ten years. You can search EOL by any of these names and find them in the names tab of any taxon page.

      Articles:
      We have articles in many languages. The article tab has a language filter, which is set to English by default. We hope soon to make that default setting configurable in your EOL profile.

      Structured data:
      Our valiant translatewiki community

      One of the great advantages of structured attribute and interaction data is that it is very efficient to translate. Commonly used data terms like "body length" and "predator" in our taxon page summaries are translated by the translatewiki community, and many place names have translations available from our geographic terms providers, geonames and wikidata. We have harvested the common names holdings of the wikidata, which include just over 291,000 names in 279 languages. We also have >93,000 common names in 130 languages added by valiant EOL members to fill gaps they observed over the past ten years. You can search EOL by any of these names and find them in the names tab of any taxon page.

      https://www.wikidata.org/wiki/Wikidata:SPARQL_query_service/queries/examples
      https://www.wikidata.org/wiki/Wikidata:SPARQL_query_service/queries/examples#The_Netherlands
      https://www.wikidata.org/wiki/Wikidata:SPARQL_query_service/queries/examples#Gender_distribution_in_the_candidates_for_the_Dutch_general_election_2017
      From Freebase to Wikidata: The Great Migration
      https://static.googleusercontent.com/media/research.google.com/nl//pubs/archive/44818.pdf
      https://www.eol.org/docs/what-is-eol/language-support
      https://upload.wikimedia.org/wikipedia/commons/4/4a/Biodiversity_Next_conference_poster_on_Wikimedia_and_iNaturalist.pdf

      https://www.wikidata.org/wiki/Wikidata:WikiProject_Freebase
      https://www.wikidata.org/wiki/Wikidata:WikiProject_Biodiversity

      iNaturalist (Q16958215) is a citizen science project focused on biodiversity. It has a large community of enthusiasts, of which some are also active in the various Wikimedia communities. This wikiproject aims at improving the cross-pollination between iNaturalist and Wikimedia communities. Wikimedia Commons is an ideal platform to source iNaturalist with observations while iNaturalist with its high-grade annotations of observations provides valuable references to Wikidata statements. "Research grade" observations are incorporated into other online databases such as Global Biodiversity Information Facility (Q1531570). iNaturalist supports many Wikimedia-compatible licensing options, including CC0 (Q6938433), Creative Commons Attribution (Q6905323) and Creative Commons Attribution-ShareAlike (Q6905942). Snippets from Wikipedia are also used on iNaturalist to describe individual taxa.
      https://upload.wikimedia.org/wikipedia/commons/4/4a/Biodiversity_Next_conference_poster_on_Wikimedia_and_iNaturalist.pdf

      EOL took 291,000 common names in 279 languages from WikiData
      EOL took 291,000 common names in 279 languages from WikiData (20)

      Publicado por ahospers mais de 3 anos antes

      Corresponding author: Ken-ichi Ueda (kueda@inaturalist.org)
      Received: 29 Sep 2020 | Published: 01 Oct 2020
      Citation: Ueda K-i (2020) An Overview of Computer Vision in iNaturalist. Biodiversity Information Science and
      Standards 4: e59133. https://doi.org/10.3897/biss.4.59133

      --

      Abstract
      iNaturalist is a social network of people who record and share observations of biodiversity.
      For several years, iNaturalist has been employing computer vision models trained on
      iNaturalist data to provide automated species identification assistance to iNaturalist
      participants. This presentation offers an overview of how we are using this technology, the
      data and tools we used to create it, challenges we have faced in its development, and
      ways we might apply it in the future.
      Presenting author
      Ken-ichi Ueda
      Presented at
      TDWG 2020
      https://www.youtube.com/watch?v=xfbabznYFV0

      Publicado por ahospers mais de 3 anos antes

      https://forum.inaturalist.org/t/identification-quality-on-inaturalist/7507

      AI Model 6 . July 2021

      The number of taxa included in the model went from almost 25,000 to over 38,000. That’s an increase of 13,000 taxa compared to the last model, which, to put in perspective, is more than the total number of bird species worldwide. The number of training photos increased from 12 million to nearly 21 million.
      Accuracy
      Accuracy outside of North America has improved noticeably in this model. We suspect this is largely due to the nearly doubling of the data driving this model in addition to recent international growth in the iNaturalist community. We’re continuing to work on developing a better framework for evaluating changes in model accuracy, especially given tradeoffs among global and regional accuracy and accuracy for specific groups of taxa.

      The recent changes removing non-nearby taxa from suggestions by default have helped reduce this global-regional accuracy tradeoff, but there’s still more work to do to improve how computer vision predictions are incorporating geographic information.
      https://www.inaturalist.org/blog/54236-new-computer-vision-model
      Participate in the annual iNaturalist challenges: Our collaborators Grant Van Horn and Oisin Mac Aodha continue to run machine learning challenges with iNaturalist data as part of the annual Computer Vision and Pattern Recognition conference. By participating you can help us all learn new techniques for improving these models.

      Start building your own model with the iNaturalist data now: If you can’t wait for the next CVPR conference, thanks to the Amazon Open Data Program you can start downloading iNaturalist data to train your own models now. Please share with us what you’ve learned by contributing to iNaturalist on Github.
      BIODIV Next
      Een presentatie hoe een be-nl model samengesteld is https://observation.org/download/Biodiv%20Next%20-%20Dutch_Belgian%20species%20ID%20.pptx
      Hierarchisch Model Ensemble is nauwkeuriger dan een singe model, mogelijk omdat bij 16.000 soorten te veel keuzes gemaakt moeten worden (Inception-v1, Inception-v3,Inception-v4, ResNet-18,, ResNet-34 , ResNet-101, GoogleLeNet, BN-NIN, GG-10)
      Performance vs Voorkomen

      https://www.inaturalist.org/blog/archives/2022/05
      https://www.inaturalist.org/blog/66531-we-ve-passed-100-000-000-verifiable-observations-on-inaturalist
      https://www.inaturalist.org/blog/63931-the-latest-computer-vision-model-updates

      AI Model 7 . May 2022

      In 2017 the amount of recognised species was 20.000 and now it is still.....20.000?
      https://www.inaturalist.org/pages/help#cv-taxa
      FWIW, there's also discussion and some additional charts at https://forum.inaturalist.org/t/psst-new-vision-model-released/10854/11
      https://forum.inaturalist.org/t/identification-quality-on-inaturalist/7507
      AI Model 4 . July 2019 included 16,000 taxa and 12 million training photos.
      AI Model 5 . July 2020 included 25,000 taxa and xx million training photos.
      AI Model 6 . July 2021 included 38,000 taxa and 21 million training photos. Training job in October 2021, we planned to train a AI Model 7 . May 2022 on 47,000 taxa and 25 million training images but finished with er 55,000 taxa and over 27 million training images.
      March 2020
      https://www.inaturalist.org/blog/31806-a-new-vision-model
      Juli 2021
      https://www.inaturalist.org/posts/54236-new-computer-vision-model
      2022
      https://www.inaturalist.org/blog/63931-the-latest-computer-vision-model-updates

      Publicado por ahospers mais de 1 ano antes

      Adicionar um Comentário

      Iniciar Sessão ou Registar-se to add comments