Progress Update – August 31st 2020 – 1723+ Zoa/Paly variations curated so far for image recognition
So I started working on this project on August 21st, 2020. Not realizing how big of a task this is going to be. This is a Zoanthid AI recognition project.
I am currently scraping online Zoanthid/Palyathoa sellers using python/selenium and will continue to do so. My current goal is to collect enough photos of different Zoanthid/Paly morphs, that I can feed them into a neural net via Tensorflow and combine that learning with image recognition in order to identify polyps by photo. I currently have a total of 1723 variations with a total of 3045 photos from 33 sources.
My main goal long term is to be able to take a photo or video of your aquarium and have the tool identify every zoa/paly variation you have in your tank.
Why are you doing this?
I’m a member of Zoanthid/Palythoa ID groups on Facebook, Reddit, and various forums. This is a tool that everyone needs in their life, the manual way of asking and waiting for a response from someone doesn’t seem to work across the board for everyone. This tool will solve that problem.
I’m having a blast so far with the image curation and can’t wait to get to that point where I can share it with the reefing community. With 1700 named variations of zoas so far (some duplicates), I’ll continue gathering photos from the web until I have amassed such a large number of photos that I can teach an AI to identify the morphs.
Online guides say I need to amass 1000 images per class/morph. I have a long way to go, but I just started and won’t give up now.
Thanks for reading.