@article{ART003338891},
author={Suhwan Yun and Semin Kim},
title={Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification},
journal={Journal of The Korea Society of Computer and Information},
issn={1598-849X},
year={2026},
volume={31},
number={5},
pages={85-93}
TY - JOUR
AU - Suhwan Yun
AU - Semin Kim
TI - Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification
JO - Journal of The Korea Society of Computer and Information
PY - 2026
VL - 31
IS - 5
PB - The Korean Society Of Computer And Information
SP - 85
EP - 93
SN - 1598-849X
AB - In medical imaging, acquiring large-scale and balanced datasets is challenging due to privacy constraints and high annotation costs, and acne severity data are typically small-scale and highly imbalanced, making accurate classification of minority severe grades particularly important. This study proposes a dual-head self-supervised learning framework tailored for small-scale, imbalanced settings. To integrate the strengths of global alignment in SimCLR and ranking optimization in AUC-CL, two projection heads are constructed in parallel on a shared encoder, where InfoNCE loss and hard negative mining–based AUC ranking loss are respectively applied. Experimental results on ACNE04 demonstrate that the proposed method outperforms conventional supervised and existing self-supervised approaches overall, with notable improvements in F1-score and significant gains in classification performance for minority severe grades.
KW - ACNE04;Self-Supervised Learning;Dual-head Architecture;Imbalanced Classification;Contrastive Learning
DO -
UR -
ER -
Suhwan Yun and Semin Kim. (2026). Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification. Journal of The Korea Society of Computer and Information, 31(5), 85-93.
Suhwan Yun and Semin Kim. 2026, "Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification", Journal of The Korea Society of Computer and Information, vol.31, no.5 pp.85-93.
Suhwan Yun, Semin Kim "Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification" Journal of The Korea Society of Computer and Information 31.5 pp.85-93 (2026) : 85.
Suhwan Yun, Semin Kim. Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification. 2026; 31(5), 85-93.
Suhwan Yun and Semin Kim. "Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification" Journal of The Korea Society of Computer and Information 31, no.5 (2026) : 85-93.
Suhwan Yun; Semin Kim. Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification. Journal of The Korea Society of Computer and Information, 31(5), 85-93.
Suhwan Yun; Semin Kim. Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification. Journal of The Korea Society of Computer and Information. 2026; 31(5) 85-93.
Suhwan Yun, Semin Kim. Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification. 2026; 31(5), 85-93.
Suhwan Yun and Semin Kim. "Dual-Head Based Contrastive Self-Supervised Learning for Imbalanced Acne Severity Classification" Journal of The Korea Society of Computer and Information 31, no.5 (2026) : 85-93.
