BEGIN:VCALENDAR VERSION:2.0 PRODID:Linklings LLC BEGIN:VTIMEZONE TZID:America/Denver X-LIC-LOCATION:America/Denver BEGIN:DAYLIGHT TZOFFSETFROM:-0700 TZOFFSETTO:-0600 TZNAME:MDT DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0600 TZOFFSETTO:-0700 TZNAME:MST DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20250522T212948Z LOCATION:Mile High Pre-Function Area DTSTART;TZID=America/Denver:20240731T090000 DTEND;TZID=America/Denver:20240731T173000 UID:siggraph_SIGGRAPH 2024_sess407@linklings.com SUMMARY:Posters: Images, Video & Computer Vision DESCRIPTION:48. T2DyVec: Leveraging Sparse Images and Controllable Text fo r Dynamic SVG\n\nWe propose T2DyVec, which leverages text prompts and spar se images as a control for dynamic vector generation. It incorporates Vect or Consistency, Semantic Tracking, and VPSD to optimize the vector paramet ers, enabling the generation of multi-frame dynamic coherent vectors. This approach can help de...\n\n\nJiakai Wu, Jun Xiao, and Haiyong Jiang (Scho ol of Artificial Intelligence, University of Chinese Academy of Sciences)\ n---------------------\n39. Crowdsourced Streetview: Integrating Real-Tim e Imagery Updates into Google Streetview\n\nCrowdsourced Streetview enhanc es Google Streetview with real-time imagery updates by leveraging crowd-so urced images from social media platforms to align and overlay these images onto existing Streetview data and provide the most current visual represe ntation of streets, landmarks, and neighborhoods...\n\n\nRyan Hardesty Lew is (Cornell Tech, University of Texas Austin)\n---------------------\n41. Image Segmentation from Shadow-Hints using Minimum Spanning Trees\n\nWe pr opose a novel image segmentation method from shadow masks. These shadow ma sks are used to detect outline points which are Delaunay triangulated. Our algorithm operates on these triangulations to close incomplete contours a nd create an image segmentation.\n\nINVITED TO THE FIRST ROUND OF THE STUD E...\n\n\nMoritz Heep and Eduard Zell (University of Bonn)\n-------------- -------\n36. Classifying Texture Anomalies at First Sight\n\nThis poster s ummarizes our recent line of research on localization and classification o f anomalies in real-world texture images. It presents our novel method for zero-shot anomaly localization (FCA), its extension to leverage contamina ted data, and anomaly clustering through contrastive learning.\n\n\nAndrei -Timotei Ardelean and Tim Weyrich (Friedrich-Alexander-Universität Erlange n-Nürnberg (FAU))\n---------------------\n42. Non-Line-of-Sight Imaging ba sed on Dual Photography using Leaked EM Waves\n\nWe present for the first time a Non-Line-of-Sight(NLoS) visualization method based on leaked elect romagnetic(EM) waves captured by antenna.\n\n\nMasaya Oishi (Chiba Univers ity), Taiki Kitazawa and Yuichi Hayashi (Nara Institute of Science and Tec hnology), and Hiroyuki Kubo (Chiba University)\n---------------------\n38. Controllable Neural Reconstruction for Autonomous Driving\n\nWe introduce an automated pipeline designed for training neural reconstruction models by leveraging sensor streams gathered from a data collection vehicle. Subs equently, our simulator, aiSim, is employed to generate a controllable vir tual counterpart of the real-world environment, enabling the repla...\n\n\ nMáté Tóth, Péter Kovács, Zoltán Bendefy, Zoltán Hortsin, and Tamás Matusz ka (aiMotive)\n---------------------\n37. Consistent Image Registration fo r Multi-view Focus Bracketing in Micro-Scale Photogrammetry\n\nWe propose a method to consistently register focus bracketing photographs taken from various viewpoints using a calibration checkerboard. Evaluation using both artificial and actual measurement datasets demonstrated that our method a chieved higher accuracy than the traditional approach.\n\n\nKodai Yasuda, Yuki Yabumoto, Takuhiro Nishida, and Takashi Ijiri (Shibaura Institute of Technology)\n---------------------\n40. Evaluating And Improving Disparity Maps Without Ground Truth\n\nOur method leverages several heuristics to a ssess the quality of a disparity map for a stereoscopic 3D image pair, wit hout needing to compare against ground truth, and can also infer disparity at pixels with unknown disparity values.\n\n\nAndreea Pocol (University o f Waterloo), Lesley Istead (Carleton University), and Craig Kaplan (Univer sity of Waterloo)\n---------------------\n49. Visualization of Flow Direct ion using Polarization Angle Changes of Cellulose Nanofiber Suspension\n\n We propose a novel flow direction visualization method using polarization angle changes of cellulose nanofiber(CNF) suspensions.\n\nINVITED TO THE F IRST ROUND OF THE STUDENT RESEARCH COMPETITION\n\n\nRyusei Okamoto (Chiba University), Shogo Yamashita and Takuya Kato (ExaWizards), and Hiroyuki Ku bo (Chiba University)\n---------------------\n45. Scribble: Auto-Generated 2D Avatars with Diverse and Inclusive Art-Direction\n\n"Scribble" automat ically generates animated 2D avatars from a selfie, stylized based on incl usive art direction. While many face stylizations are biased or inherently limit diversity, we semantically stylize specific facial features to ensu re representation. Our diversity-first art direction and sty...\n\n\nLohit Petikam, Charlie Hewitt, and Shideh Rezaeifar (Microsoft)\n-------------- -------\n34. 3DCrewCap: Applying 3D Volumetric video capture for XR helico pter rescue crew training and simulation.\n\nIn this research, we present a practical early-stage development of 3D volumetric video capture and pla yback workflow for use in helicopter rescue crew training on XR HMDs. We b reak down the workflow of using Gaussian Splat approach to construct keyfr amed 3D animated models of the rescue crew traini...\n\n\nJohn McGhee, Con an Bourke, Robert Lawther, and Hao Zhou (University of New South Wales, 3D XLab) and Rolf Petersen (Toll Helicopters (NSW) Pty Ltd)\n---------------- -----\n44. Recreating the Sodium Vapor Matting Process\n\nWe recreate the Sodium Vapor Process for compositing live-action onto a new background usi ng off-the-shelf components: two digital cinema cameras, a prism beamsplit ter, low-pressure sodium vapor lamps, and a set of dichroic bandpass and n otch reject filters.\n\n\nPaul Debevec (Eyeline Studios, USC Institute for Creative Technologies (ICT)) and Nikolas Pueringer (Corridor Digital)\n-- -------------------\n35. Adaptive Sampling for Monte-Carlo Event Imagery R endering\n\nThe novel event-based camera simulation system based on physic ally accurate Monte Carlo path tracing with adaptive path sampling based o n the probability of event occurrence.\n\n\nYuichiro Manabe (Chiba Univers ity), Tatsuya Yatagawa (Hitotsubashi University), Shigeo Morishima (Waseda University), and Hiroyuki Kubo (Chiba University)\n---------------------\ n43. Reconstructionless Airborne Radiance Fields\n\nIn this study, we pres ent a pipeline that facilitates the conversion of log files created during UAV flights to effectively harness image and sensor data to train NeRF-li ke models in only two seconds, eliminating the requirement for computation ally intensive image-based reconstructions, which take o...\n\n\nChristoph Praschl and Leopold Böss (University of Applied Sciences Upper Austria) a nd David Schedl (University of Applied Sciences Upper Austria, Hagenberg)\ n---------------------\n46. Sign Motion Generation by Motion Diffusion Mod el\n\nWe propose a method that uses a diffusion model to generate sign mot ions from the text and label prompts, enabling the generation of complex h and and body movements required to express sign words, which has been diff icult in previous work.\n\n\nKohei Hakozaki, Tomoya Murakami, Tsubasa Uchi da, Taro Miyazaki, and Hiroyuki Kaneko (Japan Broadcasting Corporation (NH K))\n---------------------\n47. Spectral Periodic Networks for Neural Rend ering\n\nWe use deep sinusoidal neural networks to represent periodic imag es in multi-resolution and to optimize non-tileable patches of textures, c reating periodic material textures from them.\n\n\nHallison Paz, Tiago Nov ello, and Luiz Velho (IMPA)\n\nRegistration Category: Full Conference, Ful l Conference Supporter, Experience, Exhibitor Full Conference, Exhibitor E xperience END:VEVENT END:VCALENDAR