The evaluation metric for this challenge is the Log Loss. Can you confirm the metric is : cross-entropy (loss) for the case of multi-class classification?
I also saw evaluation section of the challenge, that probabilities do not add to one: (.034 + .215 + .567) != 1
Image_ID blast brown healthy
id_3zpci62t81.jpg .034 .215 .567
@Zindi Can you confirm if we are dealing with multiclass or multilabel classification?
Thanks.
(Not Zindi, but:) The numbers in their examples are usually just to show the format. There is only ever ONE label per image (so multiclass) and you're right about the metric.
Thank you very much for you input.