The reason why it is difficult for an app to identify a cat breed correctly

You’ll probably find more than one app online which you can download to your smart phone (cell phone) which claims to identify cat breeds from a photograph. The picture shows a screenshot from one which is available on Google Play and it’s called Cat Scanner: Breed Recognition. It has a 4.2-star review with 1 million+ downloads and is therefore popular.

I’m going to stay right away that the 4.2 out of five review is probably exaggerated because people don’t expect it to be accurate and it isn’t accurate but they do expect it to be enjoyable to use and to have some fun with it, which the app achieves. The review comments indicate a pretty high failure rate.

App claims to be able to identify the cat breeds but how accurate is it?
Two useful tags. Click either to see the articles:- Toxic to cats | Dangers to cats

The reason why it is very difficult and sometimes impossible to distinguish one breed from another on appearance alone particularly from a photograph is because often times the breeds are not very clearly distinguished, one from the other. It is the job of the cat fancy to make sure that they are distinguishable because if they aren’t the cat fancy won’t work at all.

For instance the American Shorthair used to be a moggy and it can look like a moggy although normally more refined and ‘finished’. The Scottish Straight, Singapura, Turkish Van, Turkish Angora, Siberian, Savannah (F5), traditional Persian, Khao Manee, Havana, Chartreau, Burmilla, and Bombay call all look like random bred cats sometimes.

RELATED: Unneutered male cats have a more masculine facial appearance compared to neutered male cats

Some breeds are very distinguishable such as the Maine Coon and the flat-face Persian. The Manx cat with no tail is also going to be distinguishable using an app provided the entire cat is photographed.

However, in June 2012, a study was published in which the scientists created a model i.e. a method to distinguish breeds, both dogs and cats, from photographs. Using these models they obtain an average accuracy of about 59% which they considered to be very good “considering the difficulty of the problem”.

In the summary they state that “The visual problem is very challenging as these animals, particularly cats, are very deformable and there can be quite subtle differences between the breeds.” I believe the word “deformable” in this context means that the shapes can change. Of course they can as the cat moves in various ways including travelling or simply moving parts of their anatomy.

RELATED: Working out a domestic cat’s age from appearance

And many cat breeds have a wide range of cat coat types and the coat types are common to many cat breeds. If, let’s say in a different world, Maine Coon cats could only be bred with tabby coats then a computer could pick out a Maine Coon cat simply from their coat but as mentioned this doesn’t apply. Maine Coon cats come in a whole range of coat types and patterns which muddies the water.

Dogs are going to be easier to distinguish between the breeds because people have been breeding dogs (20k years) for about twice as long as they have been breeding cats and they were bred to be working dogs, to carry out certain functions and therefore their shape and overall appearance including their head shape sometimes differs quite substantially. As does their size by the way.

But as you can see the success rate in 2012 without using a mobile phone app was a little better than 50%. And I suspect that this app doesn’t do much better.

I have always said that it is problematic to pick out a cat breed from appearance alone. It can be done and is done and it can be done successfully many times but sometimes it can’t. Another problem was comes to mind is that cat breeders are not stamping out a product as in a factory.

There is variation between the breeders of one cat breed. If you reviewed the work of 10 breeders of British Shorthair cats, there creations wouldn’t all look identical. Breeding is not that predicable. There will be variations which adds an extra layer of variability when trying to assess a cat breed from an image.

Leave a Comment

follow it link and logo