I’m fairly new to deep learning and AI and my first proper project consists of a model taking 13 or so inputs and one colour output.

Currently, I’ve got 3 outputs (RGB) but I was wondering if I’d be better off encoding it as HSV or something similar as it feels more intuitive to me and small deviations in HSV values would be less visibly significant than deviations in RGB values. But maybe it wouldn’t make a difference to an ANN.

Additionally, I’m using the mean absolute error loss function to optimise the model. Does this sound like the best choice considering I want the predicted output to be close to the real value? Or should I be designing my own loss function, or using an different existing function?

I’ve struggled to find information on encoding colour output for a Neural Net so if anyone could point me toward an article or give me a hand I’d greatly appreciate it.

Thank you


1 Answer 1


I guess a 10% error in red is the same value to you as a 10% error in blue or green. But is the same true in HSV?

I only ask as in your domain there might be a preference in which case you could have a custom function.

If the error on each output is worth the same then Mean Average Error makes sense to me


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.