I’m exploring an AI assisted fur grading tool where you would upload a few photos and get a read on early, prime, or late depending on your region in the country. Right now I’m just testing the waters before building anything serious. Would like to hear from y’all, the good, the bad, and the ugly, and what fur bearer you would like to see started with if you would like to see it. This isn’t to replace the experts, just more of a tool for new trappers.

Joe

What kind of datasets are you planning to train it on?

Problem is that most fur needs to be felt to be graded properly. You can tell from a picture most times if the leather is prime or not and the damage.

I’ve reached out to the National Trappers Association and Fur Takers of America and planning on messaging Groenwold Fur again, and Fur Harvesters Inc. I'm not a fur grader and I wouldn't grade my own homework. I’m planning on going to the experts first before I train it. The final dataset will be built by people who know fur. Also, if you know experts who might be interested, please send me a DM.

I bet you will catch a lot of flack on here over this. But I like it. I know nothing about how it would work or if it will, but I like the innovation. Good luck to you!

Very interesting idea. It would be fun to play with a tool like that. I hope it works out for you.

It is an interesting idea but how does one build or compile a dataset of something that is almost entirely subjective ?

Much of fur grading is subjective. How would AI deal with that??

Bobcat bellies - A or B? Matter of opinion. So let’s call it an A-B. Does that really tell you anything?

Coyote weight - heavy or semi-heavy? Matter of opinion. And also section of country a pelt is from. A Texas coyote, for example, a buyer familiar with Texas coyotes will say it is a semi-heavy. A coyote buyer from the west might call it flat. So which is it.

2 dimensional pictures can’t show everything.

Interesting idea though.

Nope, can't see it working. Can't stand "everything has to be done by a computer" mentality now-a-days.

I try to preach "hands on" to all people.

Reminds me when I saw a kid at the crick one day. I knew his dad so I went over to him. He was sitting in a fold up chair with a "electronic fishing game" his dad was 50 yards upstream landing a steelhead.

I know we live in an electronic age now.
Good luck to you and your ambitions. Just 100% not me

“AI”

Artificial Intelligence and/or Augmented intelligence

In this case”fur grading”,,,, Augmented intelligence would be the route

It’s certainly possible

In the future you won't even have to wipe your own "donkey". AI will handle it.

NAFA used to have a grading manual you could download. I have a copy somewhere. Pics were drawings...not pics of actual fur, but it is informative.

I think what your trying to put together is a good thing. Pay no attention to the haters.

Sure, it aint gonna be like touching a pelt, but what fur grading manual is?

The real issue is all the rat marshes and other furbearer habitat that get filled, paved, and developed to support the data centers. AI is lame. Especially to a trapper.

AI stuff is crap,pictures voices news stories-all crap compared to when actual humans do it.
But if it helps more people to be useless all good.

There’s a bunch of fur grading info on the web, but it’s hard to find quality stuff. Books are out there, the best one being Wild Furbearer Management and Conservation in North America. It’s got everything you need to start, but it’s not cheap. I’ve got a couple and one will be at the auction after the Wolfers Summit banquet ID this August.
Interesting idea, it'd get real interesting if you get a fur grader involved. It’ll be a steep climb with some big road blocks to navigate though.

It's very feasible, but it would require a ton of pictures of varying quality to imitate the kind of imagery the system is supposed to receive from end-users. Not studio-quality DSLR shots. It's something you could crowdsource if you find trappers or buyers interested in participating. Lots of annotation work, too. You'll also need experts to help you evaluate and fine-tune the results of the training.



That book is good for learning stuff yourself, but it's not the kind if data you feed to LLM models. Basically, you'll need lots of different pictures of furs with standardized descriptions of what's in each picture (essentially, grading information), and then several times as many pictures that AI can practice on.



AI is just a useful tool. A lot of people use it to create AI slop crap, but people crave crap, so it's pretty useful for them.

AI is actually quite convenient for bulk visual evaluation, and can make tedious and boring tasks much faster and easier, plus well-trained models develop an analog of a "hunch" and can spot anomalies or other notable things better and more reliably than trained humans (especially humans tired by monotonous work). One simply needs to keep in mind the limitations of its applications.