Yuna Song

Ms. Yuna Song

orcid

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
26 November 2025 in IEEE Access
Verified
Journal Article
Open access
규이규현 이
규현 이
JHJihoon Hong
Jihoon Hong
BKBeom Seok Kim
Beom Seok Kim
YSYuna Song
Yuna Song
DLDong-Hee Lee
Dong-Hee Lee
The practical deployment of object detectors is often hindered by a gap between standard academic metrics and real-world requirements—namely, performance at a specific operating threshold, cost-asymmetric errors, and performance stability under distribution shifts. To bridge this gap, we introduce the Cost-Stability Evaluation Framework (CSEF), a systematic workflow that transforms the selection of a model and its operating threshold from a heuristic practice to a rational, data-driven decision process. CSEF operationalizes user-defined requirements—an IoU threshold (τ) for localization accuracy; a cost asymmetry parameter (β) that serves as the recall-to-precision weight in the F-beta score; and a maximum tolerable performance degradation (σreq)—to algorithmically recommend the most operationally reliable model–threshold pair. CSEF first selects the threshold that maximizes the F-beta score on the validation set, and then quantifies robustness as the relative change in the F-beta score, σdeg, from the validation set to an unseen held-out test set. Candidates that pass a conservative stability gate—where the bootstrap upper bound of |σdeg| does not exceed σreq—are then ranked by their operational F-beta score at the fixed threshold. CSEF’s utility is demonstrated across four industrial scenarios, from medical imaging to semiconductor manufacturing, highlighting its adaptability to domain-specific requirements. Ultimately, CSEF provides a practical and auditable methodology for deploying object detectors with predictable performance and reliability, thereby helping close the long-standing gap between academic evaluation and industrial deployment.

19 May 2026 in Algorithms
Verified
Journal Article
Open access
PDF file
규이규현 이
규현 이
JHJihoon Hong
Jihoon Hong
BKBeom Seok Kim
Beom Seok Kim
YSYuna Song
Yuna Song
DLDong-Hee Lee
Dong-Hee Lee
Object detection is a core component of industrial vision systems in manufacturing, infrastructure monitoring, and safety-critical sensing. While the mean average precision (mAP) averages the performance over all confidence thresholds, real-world deployment demands committing to a single operating threshold under score imprecision, distribution shifts, and asymmetric—often only approximately known—error costs. From a soft-computing perspective, deployment should explicitly manage this uncertainty rather than rely on a static validation optimum. We propose domain-specific and robust localization recall precision (DSR-LRP), a three-phase decision-support framework. The framework elicits soft domain preferences—such as asymmetric error costs, tolerable localization imprecision, and expected perturbations—from practitioner knowledge and encodes them as three quantitative parameters (k, αIoU, β). A cost-sensitive, threshold-local objective aggregates the performance within a robustness band around each candidate threshold, jointly capturing the accuracy and local stability. Finally, it yields an interpretable recommendation package comprising the operating threshold, its DSR-LRP score, and visual evidence. Experiments on four practical datasets (blood cell screening, wildfire smoke monitoring, pothole detection, and semiconductor sensor inspection) showed that DSR-LRP consistently selected operating thresholds that were robust and cost-aligned. For example, in pothole detection, an LRP-optimal threshold degraded by 15.6% under simulated shifts, while the DSR-LRP recommendation changed by only 1.8%. DSR-LRP complements global metrics such as the mAP and provides a soft-computing-oriented tool for reliable, evidence-driven deployment of industrial object detectors.