<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Selective Calibration | MacsLAB</title><link>https://jbnu.macs.or.kr/ko/tag/selective-calibration/</link><atom:link href="https://jbnu.macs.or.kr/ko/tag/selective-calibration/index.xml" rel="self" type="application/rss+xml"/><description>Selective Calibration</description><generator>Hugo Blox Builder (https://hugoblox.com)</generator><language>ko</language><lastBuildDate>Fri, 01 Jan 2027 00:00:00 +0900</lastBuildDate><image><url>https://jbnu.macs.or.kr/media/icon_hu006eacee63deb1d8999057a3cbfdb748_78083_512x512_fill_lanczos_center_3.png</url><title>Selective Calibration</title><link>https://jbnu.macs.or.kr/ko/tag/selective-calibration/</link></image><item><title>Learning When to Correct: Selective Calibration for Monocular Kinematics</title><link>https://jbnu.macs.or.kr/ko/publication/0048-learning-when-to-correct-selective-calibration-for-monocular-kinematics/</link><pubDate>Fri, 01 Jan 2027 00:00:00 +0900</pubDate><guid>https://jbnu.macs.or.kr/ko/publication/0048-learning-when-to-correct-selective-calibration-for-monocular-kinematics/</guid><description>&lt;p>&lt;strong>이준호 학부연구생이 제1저자로 수행한 연구&lt;/strong>로, &lt;strong>WACV 2027 Main Conference (Application Track)&lt;/strong>에 채택되었습니다. 저자는 이준호, 강동준, 이경수입니다.&lt;/p>
&lt;p>단안 영상 기반 관절 운동학 추정에서 이미 정확한 추정값까지 보정해 오히려 오차를 키우는 과보정 문제를 다룹니다. 제안한 선택적 보정 기법은 기존 추정값을 유지할지 ExtraTrees 보정값을 채택할지 판단하여, 평균 절대 오차를 &lt;strong>5.94°&lt;/strong>로 줄이고 5° 허용 범위 내 추정값의 보존 실패율을 &lt;strong>34.7%에서 16.5%&lt;/strong>로 낮췄습니다.&lt;/p>
&lt;h3 id="선택적-보정-프레임워크">선택적 보정 프레임워크&lt;/h3>
&lt;p>&lt;a href="featured.png">
&lt;figure >
&lt;div class="d-flex justify-content-center">
&lt;div class="w-100" >&lt;img alt="기존 추정값과 ExtraTrees 보정값 사이에서 선택하는 보존 중심 선택적 보정 프레임워크." srcset="
/ko/publication/0048-learning-when-to-correct-selective-calibration-for-monocular-kinematics/featured_hu64f687059b9797a5e9271f757cae2aa4_338023_2a0ab38496cad78818d4be9fc57934aa.webp 400w,
/ko/publication/0048-learning-when-to-correct-selective-calibration-for-monocular-kinematics/featured_hu64f687059b9797a5e9271f757cae2aa4_338023_01ae6042fb505e5a6d1fb76fb02c9c53.webp 760w,
/ko/publication/0048-learning-when-to-correct-selective-calibration-for-monocular-kinematics/featured_hu64f687059b9797a5e9271f757cae2aa4_338023_1200x1200_fit_q75_h2_lanczos_3.webp 1200w"
src="https://jbnu.macs.or.kr/ko/publication/0048-learning-when-to-correct-selective-calibration-for-monocular-kinematics/featured_hu64f687059b9797a5e9271f757cae2aa4_338023_2a0ab38496cad78818d4be9fc57934aa.webp"
width="760"
height="293"
loading="lazy" data-zoomable />&lt;/div>
&lt;/div>&lt;/figure>
&lt;/a>&lt;/p>
&lt;p>&lt;em>논문 Figure 2. 피험자 단위로 분리된 학습과 forest consensus 표현을 이용해 보정의 이득을 예측하며, 추론 시 참조 측정값 없이 보정 여부를 결정합니다.&lt;/em>&lt;/p></description></item></channel></rss>