https://jbnu.macs.or.kr/en/tag/causal-compositional-graph-parsing/Causal-Compositional Graph ParsingHugo Blox Builder (https://hugoblox.com)en-usTue, 22 Sep 2026 00:00:00 +0900https://jbnu.macs.or.kr/media/icon_hu006eacee63deb1d8999057a3cbfdb748_78083_512x512_fill_lanczos_center_3.pnghttps://jbnu.macs.or.kr/en/tag/causal-compositional-graph-parsing/https://jbnu.macs.or.kr/en/publication/0046-sfxgraph-causal-compositional-graph-learning-for-stylized-sound-effects-in-webtoons/Tue, 22 Sep 2026 00:00:00 +0900https://jbnu.macs.or.kr/en/publication/0046-sfxgraph-causal-compositional-graph-learning-for-stylized-sound-effects-in-webtoons/<h3 id="framework-overview">Framework Overview</h3> <p><a href="featured.png"> <figure > <div class="d-flex justify-content-center"> <div class="w-100" ><img alt="SFXGraph framework showing SFX detection, graph parsing, text decoding, graph-conditioned rendering, and synthetic supervision." srcset=" /en/publication/0046-sfxgraph-causal-compositional-graph-learning-for-stylized-sound-effects-in-webtoons/featured_hub9034ff933b54c97d71d40d056aed2f9_1310892_06dc719057534a9dbf4858ffc16f3fd6.webp 400w, /en/publication/0046-sfxgraph-causal-compositional-graph-learning-for-stylized-sound-effects-in-webtoons/featured_hub9034ff933b54c97d71d40d056aed2f9_1310892_0fc85dfabfd1145c3c002901743f42ab.webp 760w, /en/publication/0046-sfxgraph-causal-compositional-graph-learning-for-stylized-sound-effects-in-webtoons/featured_hub9034ff933b54c97d71d40d056aed2f9_1310892_1200x1200_fit_q75_h2_lanczos_3.webp 1200w" src="https://jbnu.macs.or.kr/en/publication/0046-sfxgraph-causal-compositional-graph-learning-for-stylized-sound-effects-in-webtoons/featured_hub9034ff933b54c97d71d40d056aed2f9_1310892_06dc719057534a9dbf4858ffc16f3fd6.webp" width="760" height="510" loading="lazy" data-zoomable /></div> </div></figure> </a></p> <p><em>Figure 3 from the paper. SFXGraph represents stylized sound effects as structured graphs and uses graph-conditioned rendering and synthetic supervision to support recognition.</em></p> <h3 id="recognition-results">Recognition Results</h3> <p><a href="recognition-results.png"> <figure > <div class="d-flex justify-content-center"> <div class="w-100" ><img alt="Comparison of the original webtoon panel, four OCR baselines, and SFXGraph predictions." srcset=" /en/publication/0046-sfxgraph-causal-compositional-graph-learning-for-stylized-sound-effects-in-webtoons/recognition-results_hu9a580c52677f521afac54e46fc65adfc_297426_66e84e8fb5f296b3319fa4fb243c2bda.webp 400w, /en/publication/0046-sfxgraph-causal-compositional-graph-learning-for-stylized-sound-effects-in-webtoons/recognition-results_hu9a580c52677f521afac54e46fc65adfc_297426_f5404d672a5c8e9b5c1376bfbd32d122.webp 760w, /en/publication/0046-sfxgraph-causal-compositional-graph-learning-for-stylized-sound-effects-in-webtoons/recognition-results_hu9a580c52677f521afac54e46fc65adfc_297426_1200x1200_fit_q75_h2_lanczos_3.webp 1200w" src="https://jbnu.macs.or.kr/en/publication/0046-sfxgraph-causal-compositional-graph-learning-for-stylized-sound-effects-in-webtoons/recognition-results_hu9a580c52677f521afac54e46fc65adfc_297426_66e84e8fb5f296b3319fa4fb243c2bda.webp" width="760" height="227" loading="lazy" data-zoomable /></div> </div></figure> </a></p> <p><em>Figure 4 from the paper. From left to right: original panel, Naver Clova OCR, Keras OCR, Paddle OCR, Google OCR, and SFXGraph. Bounding boxes and transcriptions show detection and recognition outputs; &ldquo;None&rdquo; indicates a missed detection.</em></p>