How do I try it?

Currently, there is a limited way to give it a try. I hope to have more ways to use it in the future, but it currently works as a Telegram bot.

Telegram is an instant messenger app that is popular with furries. It supports private messages and groups, as well as a bot interface.

After downloading the Telegram client, the user can message the bot at @RainRatBot like they would message any other user. Unlike a user, the only things it will respond to are photos, and a few specifically formatted commands, like "/start"

What's going on here? Well the user sends their photo of this orange fursuit, using the usual Telegram features like they would send a photo to any other user.

Behind the scenes, my server downloads the message from Telegram, then analyses it uses a previously-created Machine Learning model. Then it replies:

Fursuit Name:
Sunny Dingo (Made Fur You) 97.89%

"Fursuit Name:" is included in all replies to photos.

"Sunny Dingo (Made Fur You)" is the name of the fursuit. It knows this information because earlier, photos of this same fursuit were labelled "Sunny Dingo (Made Fur You)" in the training set. (Made Fur You is a fursuit making company. All the fursuit names are labelled with their maker as well, first to give the maker proper credit, second because many fursuits have simple names, and it helps narrow down exactly which suit is being referred to.)

"97.89%" is a confidence level. Based on the information the model is created with, it is 97.89% certain that the label Sunny Dingo (Made Fur You) is correct.

You can give it a try with your own photos. Since you're going to the effort of sending photos, I programmed it to always reply to you with its best prediction, even if the confidence is very low. For instance, if you send a fursuiter that it doesn't know, it might reply with a guess of a fursuiter that it does know, but with a confidence of around 20%. If someone were using the results, they might only use results that were 80%, 90% or even higher confidence, but for this usage, since there's no harm in a wrong guess, it will always give you the result.