본 연구는 디지털 기술의 확산이 전통적 조각 개념을 어떻게 재구성하고 확장하는지를 탐구한다. 전통적인 조각이 물질성, 형식, 그리고 물리적 공간의 점유를 중심으로 정의되 어 왔다면, 현대 미술에서는 데이터와 알고리즘을 새로운 형성 재료이자 방법론으로 수용 함으로써 조각 개념에 근본적인 전환이 시도되고 있다. 본 논문은 디지털 환경에서 조각이 구현되는 구체적인 양상을 살펴보기 위해 3D 모델링, 증강현실(AR), 인공지능(AI)이라는 세 가지 핵심 기술을 중심으로 세 명의 작가 사례를 분석한다. 이를 통해 비물질적 요소들 이 현대 조각 실천에서 어떠한 방식으로 작동하는지를 고찰하고, 기술 융합의 시대에 조각 매체가 지니는 미학적 의미와 가능성을 조명하고자 한다.

자기공명영상(MRI)은 연부조직의 대조도가 우수하여 신경 및 근골격계 이상을 평가하는 데 탁월한 진단 도구로 활 용되고 있다. 특히 요천추 신경총과 같은 복잡한 신경 구조의 정밀한 영상화에 적합하여 임상 진단 과정에서 중요한 역할을 담당한다. 기존에 사용된 SPACE 3D, STIR, 그리고 최근에는 딥러닝 재구성 기법 등 다양한 MRI 기법이 도입되어 영상 화질 향상과 검사 효율성을 동시에 개선하고 있다. 본 연구에서는 조영제 주입 후 요천추 신경총 MRI 검사에서 얇은 슬라이스 두께로 획득한 SPACE 3D T2 STIR 기법과 딥러닝 TSE T2 STIR 기법을 비교하여, 요천추 신경총 영상에 유용한 MRI 기법을 알아보고자 하였다. 요천추 신경총 병변이 의심되는 20명의 환자를 대상으로 하 여 SPACE 3D T2 STIR 기법과 DL TSE T2 STIR 기법을 적용해 관상면 영상을 획득한 후, 신호대잡음비, 대조대 잡음비, 검사 시간 및 영상의 질을 정량적·정성적 방법으로 분석하였다. 그 결과, 기법 간 통계적으로 유의한 차이 가 확인되었으며, 특히 DL 기법은 검사 시간 단축과 우수한 대조도를 제공하였다. 환자의 임상 상태와 촬영 여건을 고려하여 적절한 기법을 선택한다면, 요천추 신경총 MRI 진단에 최적화된 영상 품질을 확보할 수 있을 것으로 사료 된다.

Pavement surface texture plays an essential role in skid resistance, tire–pavement interaction, and wet-weather driving safety. Conventional evaluations of pavement texture are primarily based on two-dimensional (2D) profile measurements, such as Mean Texture Depth (MTD) and Mean Profile Depth (MPD), which cannot fully represent the complex three-dimensional (3D) characteristics of pavement surfaces. As a result, the relationship between surface texture and wet pavement friction is often insufficiently described. Recent advances in high-resolution optical scanning enable detailed acquisition of surface topography and provide opportunities for more accurate texture quantification. This study proposes a comprehensive framework for characterizing micro- and macro-surface textures using high-resolution 3D scanning combined with both 2D and 3D analytical methods. Dense point-cloud data were collected from concrete pavement surfaces, and multiple longitudinal and transverse profiles were extracted. Fast Fourier Transform (FFT)-based filtering was applied to separate micro- and macro-texture components, and representative texture parameters were calculated from both profile-based and surface-based analyses. Wet pavement friction was evaluated using the British Pendulum Number (BPN), and statistical relationships between texture parameters and friction were examined. The results demonstrate that the proposed approach effectively captures multi-scale texture features and provides improved correlation with wet friction compared with traditional 2D methods. The developed methodology offers a practical basis for texture-based friction evaluation and pavement safety assessment.

This study examines whether a 3D virtual fitting system can function as an efficient digital pattern-making method for women’s jeans to complement or replace conventional 2D measurement-based design. It further explores the potential application of virtual fitting-based pattern design processes in digital fashion education and the apparel industry. To develop 3D virtual fitting-based pants, a close-fitting 3D pants pattern was generated by flattening an avatar’s lower-body surface. Initial evaluations necessitated adjustments in dart volume and placement for both the front and back panels. Stress analysis further revealed high concentrations near the crotch, requiring modifications to the crotch extension. The pattern was iteratively refined using real-time feedback from appearance changes, stress distribution, and fitting errors during virtual fitting. Post-modification evaluation results showed significant improvements across all appearance evaluation categories. Notable enhancements were found in key fit factors, including dart position and length, crotch appearance, and hip fit. Subsequently, a digital jeans pattern was designed based on the refined close-fitting 3D pants pattern. Stepwise modifications informed by virtual fitting data led to improvements in both appearance and silhouette completeness. Comparative evaluation of jeans produced using the proposed 3D-derived pattern and a conventional 2D pattern showed no significant differences across most assessment items. However, the 3D-derived pattern scored significantly higher in critical areas such as thigh appearance and knee-line positioning. This indicates that 3D-based pattern design is particularly effective in areas requiring accurate reflection of 3D body curvature and movement characteristics.

최근 메타버스 기술의 확산으로, 2D 도면을 넘어 3D 공간을 직접 시뮬레이션하려는 실내 디자인 서비스 수요가 급증하고 있다. 그러나 기존 3D 재구성 기술은 높은 연산량에 따른 긴 처리 시간과 실제 공간과의 시각적 괴리, 그리고 재구성 후 수정이 어려워 실시간 서비 스 적용에 한계가 있다. 본 논문에서는 이러한 문제를 해결하기 위해 기존 PE3R 모델을 획 기적으로 개선한 IN3R(Interior New 3D Reconstruction) 프레임워크를 제안한다. IN3R은 YOLOE 기반의 단일 통합 파이프라인을 도입하여 객체 탐지, 분할, 의미 부여 과정을 경량 화함으로써 불필요한 연산을 제거하고 재구성 속도를 최적화하였다. 또한, 생성형 LLM을 결합하여 공간의 맥락을 분석하고, 사용자가 텍스트 프롬프트를 통해 가구 배치와 분위기 를 능동적으로 편집할 수 있는 인터랙티브 기능을 구현하였다. 실험 결과, 제안 모델은 기 존 PE3R 대비 재구성 품질의 저하 없이 평균 3.65배 빠른 처리 속도를 달성하였다. 본 프 레임워크는 실내 재구성 기술의 실용성을 확보하여, 향후 사용자 맞춤형 가상 인테리어 및 이커머스 연계형 3D 쇼룸 시스템의 핵심 기술로 활용될 것으로 기대된다.

This study examines the appearance characteristics of collars produced through fused deposition modeling (FDM) 3D printing by analyzing the interaction between collar types and infill structures. Three collar types-stand, convertible, and shirt-and three infill structures-rectilinear, gyroid, and honeycomb-were selected to generate nine collar samples. All samples were modeled in CLO 2025 with identical base dimensions and printed under consistent FDM settings using TPU filament to ensure comparable structural conditions. Appearance evaluations were conducted by 12 experts in garment construction and fashion design through an online survey on the naturalness and visual quality of each collar using a 5-point Likert scale. The findings revealed that the optimal infill structure varied depending on collar type. The stand collar exhibited stable appearance evaluations across all infill structures, with the gyroid structure receiving the highest evaluations. The convertible collar showed the greatest sensitivity to infill structure, with the rectilinear infill structure producing the highest ratings for both naturalness and visual quality. The shirt collar demonstrated the lowest scores overall due to its structural complexity; however, the honeycomb structure was relatively more suitable. ANOVA results confirmed statistically significant differences among collar types for the rectilinear and gyroid structures, but not the honeycomb structure. This study provides empirical evidence that internal geometry plays a critical role in determining the silhouette quality of 3D-printed garment components. The results offer practical guidelines for selecting infill structures tailored to specific collar shapes and contribute foundational insights toward the development of modular, seamless garment systems.

This study evaluated the fit of a one-piece dress using a 3D-printed dress form designed to reflect the body shape of middle-aged women and examined its potential for practical application by comparing the results with those from a 3D virtual fitting program. Therefore, a dress form was created based on the body measurements of middle-aged women using 3D body scanning and 3D printing, and an actual one-piece dress was fitted onto it. The same pattern was then simulated in a 3D virtual fitting program. A subsequent visual assessment was performed to compare and analyze the similarity between the 3D virtual fitting and actual fitting results. The analysis revealed that the 3D-printed dress form more accurately replicated the body characteristics of middle-aged women, making it advantageous for evaluating actual wearing comfort and garment fit. In contrast, the virtual fitting program demonstrated limitations in detailed expressions of elements such as wrinkles in specific body areas and fabric properties, resulting in lower consistency with real-world fitting outcomes compared to the dress form. This study confirmed that the 3D-printed dress form for middle-aged women can enhance accuracy in both fit evaluation and garment production processes. Future studies should focus on developing dress forms that accommodate diverse body types and refining virtual fitting technologies to enable more precise garment simulation and evaluation.

조선해양산업의 용접 자동화는 숙련 인력 부족과 고위험 환경 극복을 위해 협동로봇 중심으로 발전하고 있으나 선박블록 내 부 공간을 계측하기 위한 3차원 비전 센서의 경우 계측 거리에 따라 품질이 저하되는 문제가 있다. 본 연구는 협동로봇 용접 자동화를 위 해 3차원 포인트 클라우드 기반의 V-개선 용접선 검출 알고리즘을 제안하였으며 특히 자동화 용접에 필수적인 1m 미만 근거리에서 비전 센서 기술에 따른 계측 정밀도 및 검출된 평면의 개선각을 정량적으로 비교 검증하였다. 대중적으로 활용되고 있는 Active IR Stereo와 Time-of-Flight(ToF) LiDAR 센서를 400mm, 600mm, 800mm 거리에서 90° 개선각 시편으로 비교 평가한 결과 Active IR Stereo 센서는 삼각 측 량 원리의 한계로 인한 데이터 왜곡으로 상당한 각도 오차를 보였으며 800mm에서는 개선면 검출에 실패하였다. 반면 ToF LiDAR 센서는 데이터 왜곡에 강건하여 400mm에서 4.4°의 가장 낮은 평균 개선각 오차를 기록했으며 모든 거리에서 안정적으로 평면을 검출하였다. 이 를 통해 근거리 V-개선 형상 계측에는 ToF LiDAR 방식이 Active IR Stereo 방식보다 높은 정밀도를 제공하여 용접선 검출에 더 적합함을 정량적으로 검증하였다.

Recent advances in 3D data-driven digital twin research have revealed limitations in existing tree reconstruction methods, which rely solely on either scanning or procedural generation. To address this issue, this study proposes a hybrid pipeline that integrates data-driven reconstruction and procedural generation using Gaussian Splatting(GS) data. The proposed method converts multi-view GS outputs into dense point clouds and extracts a stable skeletal structure through color-density-based graph analysis. Fine branches and leaves are procedurally generated using a space colonization algorithm that incorporates botanical principles, achieving a natural and structurally coherent form. Quantitative evaluations using Chamfer distance and Intersection-over-Union metrics demonstrate high geometric similarity and volumetric consistency with the original GS data. The proposed GS-based hybrid framework ensures both visual realism and biological plausibility, enabling efficient and reliable digital twin tree modeling.

This study compares and analyzes the virtual fitting results of tailored jacket according to differences in digital production workflows using the CLO 3D virtual fitting program. The workflows presented in educational materials were examined based on the preparing, modeling, and finishing stages. This approach enabled identifying essential workflows and performing virtual fitting simulations using the same tailored jacket design to compare differences in results. The findings indicate that the pattern composition method directly influenced 3D appearance and structural stability of the garments. When more complex internal structures were incorporated, the silhouette appeared more realistic, although the increased number of contact points reduced simulation stability. Moreover, details such as armholes, sleeve slits, and welt pockets were most effectively represented in workflows that closely reflected actual sewing processes, thereby enhancing the overall completeness of the virtual garments. These results highlight the importance of choosing an appropriate workflow when considering the design objectives in digital garment production. In particular, adopting workflows based on real sewing structures is essential when the goal is to improve detail representation and overall garment completeness. This study emphasizes the academic significance of systematically examining workflow differences in CLO 3D, providing new insights into how digital garment production workflows influence simulation accuracy and quality.

Lighting plays a fundamental role in defining the visual atmosphere, emotional tone, and realism of animation. Beyond simple illumination, it functions as a visual language that conveys narrative intention and guides audience perception. In stop-motion animation, the use of miniature sets produces a distinctive “small set effect,” where lighting appears denser, sharper, and more dramatic due to the physical compression of space and reduced light distance. This phenomenon has long been recognized by practitioners as a defining aspect of the medium’s aesthetic identity.In contrast, 3D animation employing Physically Based Rendering (PBR) adheres to the inverse square law of light behavior, ensuring that when distance and intensity are proportionally adjusted, lighting results remain physically consistent across different scales. This raises a critical question: if the physical outputs are identical, why do smaller digital scenes often appear brighter or more intense to viewers?To explore this discrepancy, this study replicated the same scene at 1×, 2×, and 4× scales using Autodesk Maya and the Arnold renderer under identical lighting setups. Quantitative analysis through Difference operations confirmed pixel-level equivalence among all scales, validating the mathematical consistency of PBR. However, perceptual evaluations using 5-point Likert scales revealed a consistent bias: smaller scenes were perceived as exhibiting stronger illumination, higher contrast, and a more dramatic lighting mood.These results indicate that the “small set effect” in digital environments arises not from physical variation but from perceptual interpretations influenced by contextual visual cues. The study concludes by proposing a perceptual lighting design framework that integrates factors such as texture density, focal length, depth of field, and reference scale—bridging the gap between physical accuracy and emotional realism in digital lighting design.

This study proposes a mobile-based lightweight deep learning model (Lite-MCC) capable of reconstructing three-dimensional (3D) spatial structures from a single RGB image. Conventional 3D reconstruction models require multi-view inputs or point cloud data and depend on large-scale computational resources, which limits their real-time applicability in practical environments. To address this limitation, the proposed Lite-MCC model simplifies the existing Multiview Compressive Coding (MCC) architecture, enabling accurate 3D reconstruction using only a single image. The model adopts a parallel structure consisting of a Vision Transformer (ViT-Tiny) and a Geometry Encoder to extract visual and spatial features simultaneously, while a Transformer Decoder generates the corresponding 3D point cloud. Furthermore, depth map–based input transformation and ONNX-based optimization are employed to achieve efficient real-time inference on edge devices. Experimental results on the CO3D dataset demonstrate that Lite-MCC reduces computational cost by 87% and memory usage by 65%, while maintaining a Chamfer Distance of 0.045, comparable to the original MCC model. These results indicate that the proposed method provides a promising direction for lightweight AI models enabling low-cost, real-time 3D recording and visualization.