Question about evaluation

Hello,

Here are some question I have about the evaluation metric which is not entirely clear to me. https://justraigs.grand-challenge.org/evaluation-metrics/

If the final score is Sfinal = (R_ref + R_just)/2. and R_just is a hamming distance. R_ref is measuring the sensitivity, so the higher the better. R_just is the hamming distance, the lower the better. I don't understand why averaging these two values could indicate a better performance of the model. Could you elaborate on that please?

Also, is R_ref in percentage or float value between 0 and 1?

Thanks in advance for your response.

Hi Alirmo,

You are almost right! R_ref is not the sensitivity itself, but rather the rank of your sensitivity performance. The same applies to R_just. Those two ranks are averaged and result in a score Sfinal. Those scores are again ranked.

I hope this helps!

Best, Koen