NEWS: > 1000 records! Mosquito Conference and RISE of the MACHINES

Hi iNat Mosquito people!

I am at the American Mosquito Control Assocation annual meeting in Orlando, Florida. I just finished talking about the mosquitoes in Hawai`i project and all your great work!

My presentation entitled: "Engaging citizen scientists in mosquito response using the iNaturalist social network, app and machine learning tools" got a lot of interest at this conference.

My talk introduced your amazing work and showed off how iNaturalist now has operational Computer Vision--otherwise known as machine learning or sometimes referred to as artificial intelligence. The Computer Vision algorithm can now successfully identify several species of mosquitoes that you have documented including Aedes aegypti, Aedes albopictus and Culex quinqefasciatus.

The vision algorithm does not in any way replace the role of identifiers in iNaturalist. In fact the 'social' aspect of identification builds up the database of research grade records allowing a 'deep learning' algorithm to figure out how to recognize a species in the database.

For more information see this iNat blog post: iNaturalist Computer Vision Explorations page

ONE MORE THING: YOU'VE done it, collected >1000 mosquitoes, over >800 research grade! This is fantastic!

Over 400 Aedes albopictus almost 250 Culex quinquefasciatus and > 100 Aedes aegypti! See this link for the project observations breakdown

This is critical because once we have > 20 research grade records of a single species the algorithm starts to train, and once there are >100 it starts to work really well. So YOUR WORK is why people around the world are getting positive computer identifications of both Aedes aegypti and Aedes albopictus! THIS IS THE FIRST OPERATIONAL SYSTEM OF THIS TYPE THAT IDs THESE TWO VECTOR SPECIES, YOU SHOULD BE PROUD!

One goal of the conference is to recruit MORE people to help build the OBSERVATIONS and IDENTIFICATIONS of new species to continue training the machine learning algorithm to keep virtuous cycle going!

So thanks again for continuing to OBSERVE and also IDENTIFY mosquitoes in Hawai`i BEYOND.

All the best,

Durrell

Orlando, FL
28-Feb-2019

Posted on 01 de março de 2019, 07:24 PM by cydno cydno

Comentários

Aloha Durrell,

Thanks for the post about the vision algorithm.

I think this is a game changer........

not because it will replace the need for humans to properly ID an observation, but because it gives non-trained observers access to a "mentor" to help them learn to identify what they see. I tried the system with dozens of photographs of plants, animals, fish and was amazed at how well it worked. I showed it things I did not know, and it reliably (>9 out of 10 times) gave me what looks like the correct ID.

I hope iNat is busy integrating this into the apps and online site.

And hey, 1000 mosquito IDs in Hawaii is pretty neat!

Van

Publicado por vaneden cerca de 5 anos antes

Adicionar um Comentário

Iniciar Sessão ou Registar-se to add comments