Two Papers Accepted to ACCV 2026
We are pleased to announce that two papers from our laboratory have been accepted to the Asian Conference on Computer Vision 2026 (ACCV 2026), a Top BK/CS conference. SFXGraph: Causal-Compositional Graph Learning for Stylized Sound Effects in Webtoons Authors: Sakang Hong, Haeyun Lee, Gunha Hong, Jinyoung Jung, Sanga Ahn, and Kyungsu Lee (corresponding author).
We are pleased to announce that two papers from our laboratory have been accepted to the Asian Conference on Computer Vision 2026 (ACCV 2026), a Top BK/CS conference.
SFXGraph: Causal-Compositional Graph Learning for Stylized Sound Effects in Webtoons
Authors: Sakang Hong, Haeyun Lee, Gunha Hong, Jinyoung Jung, Sanga Ahn, and Kyungsu Lee (corresponding author).
This work introduces SFXGraph, a causal-compositional graph parsing framework for recognizing stylized sound effects in webtoons. The framework represents linguistic content, geometric deformation, visual style, and contextual interactions as a structured graph. Counterfactual factor consistency and a graph-supervised synthetic generation pipeline support interpretable representations and improved robustness across artistic styles.
SFXGraph framework overview (Figure 3 from the paper).
Biomarker-Aware Mixture-of-Experts for Device-Invariant Personalized Skin Diagnosis
Authors: Seo-Yeon Choi and Kyungsu Lee (corresponding author).
This work presents an end-to-end framework for device-invariant, personalized skin diagnosis. It combines hierarchical hybrid normalization, a keypoint-guided Skin Mixture-of-Experts, and causal expert deliberation to address device heterogeneity, illumination variability, and interactions among dermatological biomarkers.
Biomarker-aware skin diagnosis framework (Figure 1 from the paper).
Congratulations to all authors on these contributions to webtoon understanding and personalized skin diagnosis.