Class distance weighted cross entropy loss for classification of disease severity

Gorkem Polat, Ümit Mert Çağlar, Alptekin Temizel

Abstract: Highlights•Introduced a new loss function for ordinal image classification, CDW-CE.•Outperforms other methods, showcasing promise for diverse image classification tasks.•Demonstrates better feature extraction capabilities than other loss functions.•Domain experts find CDW-CE’s attention regions closely align with their insights.

External IDs:doi:10.1016/j.eswa.2024.126372