Error: Input buffer contains unsupported image format

[I. Monetaria]

Hello!

I was trying to let my model learn from added training data. I attached my source code. Month ago it worked perfectly, but now I always get the following error: 

"Error: Input buffer contains unsupported image format". 

When I go to the "Train"-section (backend) where you can add manually images, appears now for every image that couldn't be loaded a red warning sign (foto attached).

Thanks for your help!

I.

[Dale Lane]

Can you share how you're setting the you variable, please?

Kind regards

D

[I. Monetaria]

Yes, of course.

That was a default code, I changed nothing... 

[Dale Lane]

Thanks for sharing that - I can see what the problem is now.

You have this code:

  set you to (  recognise image ( costume image ) )

The costume image block returns the contents of the image

The recognise image block returns the results from using your machine learning model to classify something

Combining these into 

    recognise image ( costume image )

returns the result from classifying the costume image

In other words, this will return the string "Stein", "Papier", or "Schere".

You aren't adding image data to your training buckets. You aren't adding pictures, you are adding words like "Stein" into the training buckets. 

The error about the unsupported image format is referring to the fact that the words you are adding to the training data aren't images.

Kind regards

D

[Dale Lane]

Try changing your add training data blocks to:

add training data (costume image) (Stein/Schere/Papier)