1,129 research outputs found
Single-Image 3D human reconstruction with 3D-aware diffusion priors and facial enhancement
No full textCreating high-quality, photorealistic 3D digital humans from a single image remains challenging. While existing methods can generate visually appealing multi-view outputs, they often suffer from inconsistencies in viewpoints and camera poses, resulting in suboptimal 3D reconstructions with reduced realism. Furthermore, most approaches focus on body generation while overlooking facial consistency – a perceptually critical issue caused by the fact that the face occupies only a small area in a full-body image (e.g., ∼ 80 × 80 pixels out of a 512 × 512 image). This limited resolution and low weight for the facial regions during optimization leads to insufficient facial details and inconsistent facial identity features across multiple views.To address these challenges, we leverage the powerful capabilities of 2D video diffusion models for consistent multi-view RGB and Normal human image generation, combined with the 3D SMPL-X representation to enable spatial consistency and geometrical details. By fine-tuning the DiT models (HumanWan-DiTs) on realistic 3D human datasets using the LoRA technique, our method ensuresboth generalizability and 3D visual consistency on realistic multi-view human image generation. The proposed facial enhancement is integrated into 3D Gaussian optimization to enhance facial details. To further refine results, we apply super-resolution and generative priors to reduce facial blurring alongside SMPL-X parameter tuning and the assistance of generated multi-view normal images, achieving photorealistic and consistent rendering from a single image. Extensive experiments demonstrate that our approach outperforms existing methods, producing photorealistic, consistent, and fine-detailed human renderings
Supplemental material for Standard Reference Line Combined with One-Point Calibration-Free Laser-Induced Breakdown Spectroscopy (CF-LIBS) to Quantitatively Analyze Stainless and Heat Resistant Steel
No full textSupplemental material for Standard Reference Line Combined with One-Point Calibration-Free Laser-Induced Breakdown Spectroscopy (CF-LIBS) to Quantitatively Analyze Stainless and Heat Resistant Steel by Hongbo Fu, Huadong Wang, Junwei Jia, Zhibo Ni and Fengzhong Dong in Applied Spectroscopy</p
TAVE: Template-based Augmentation of Visual Effects to Human Actions in Videos
No full textWe present TAVE, a framework that allows novice users to add interesting visual effects by mimicking human actions in a given template video, in which pre-defined visual effects have already been associated with specific human actions. Our framework is mainly based on high-level features of human pose extracted from video frames, and uses low-level image features as the auxiliary information. We encode an action into a set of code sequences representing joint motion directions and use a finite state machine to recognize the action state of interest. The visual effects, possibly with occlusion masks, can be automatically transferred from the template video to a target video containing similar human actions.Pacific Graphics PostersPoster
InspireMePosing: Learn Pose and Composition from Portrait Examples
No full textSince people tend to build relationship with others by personal photography, capturing high quality photographs on mobile device has become a strong demand. We propose a portrait photography guidance system to guide user's photographing. We consider current scene image as our input and find professional photograph examples with similar aesthetic features for it. Deep residual network is introduced to gather scene classification information and represent common photograph rules by features, and random forest is adopted to establishing mapping relations between extracted features and examples. Besides, we implement our guidance system on a camera application and evaluate it by user study.Pacific Graphics Short PapersTowards Better Quality of Images/Video
Shi nei 45 ji he zi hao mi bo tong xin wu xian tong dao ce liang yu jian mo
No full textWang, Hongbo.Thesis Ph.D. Chinese University of Hong Kong 2015.Includes bibliographical references (leaves 215-228).Abstracts also in Chinese.Title from PDF title page (viewed on 05, December, 2016).Wang, Hongbo
Cross-layer design for reliable and efficient data transmission over multiple antenna mobile infostation networks
No full textIn this thesis, we propose a system architecture that has multiple localized networks, each with a mobile information server (Mobile Infostation) that collects the information relevant to mobile users over a back-haul network and delivers the information to the users on demand. For efficient use of the bandwidth and energy resources of the wireless network, we design and evaluate a cross-layer solution for the Radio Link Control (RLC), Medium Access Control (MAC), and Transport layers of the network protocol stack.
At the RLC layer, we propose a Hybrid-ARQ scheme with transmit power control to optimize the average energy consumption while maintaining a target packet error rate (PER) to increase the reliability of the Link Layer. We show that the above optimization problems for a short term static Rayleigh block fading MIMO channel can be formulated and solved using geometric programming. Our illustrative results show that, with a target PER of 0.01%, the optimal power allocation scheme can provide a gain of up to 3 dB for a Space-Time Trellis Code (STTC) coded MIMO channel with maximum two ARQ rounds.
At the MAC layer, we propose an efficient MAC scheme with two variations, where the data frames are scheduled for transmission based on the user priorities, channel conditions and mobility. In the reliable version of the MAC protocol named MIN-MACa, a stop-and-wait ARQ for data frames and a selective repeat ARQ for data sub-frames are combined to ensure reliable and efficient data transmission without the use of the TCP protocol. A highly efficient version of the MAC protocol without the ARQ overhead named MIN-MACb is also proposed, thereby requiring an integrated transport protocol for reliable end-to-end data transmission.
At the Transport layer, a transport protocol (MIN-TCP) optimized for the MIN-MACb is proposed to improve the end-to-end throughput. The simulation results show that, with MIN-TCP, the throughput can be doubled compared with TCP-NewReno when the PER is high. The cross-layer approach to the Transport and MAC layer co-design also provides higher throughput than the MIN-MACa approach when the PER is low.Ph.D.Includes bibliographical references (p. 108-113).by Hongbo Li
Polymer Adhesion, Friction, and Lubrication
No full textSpecifically dedicated to polymer and biopolymer systems, Polymer Adhesion, Friction, and Lubrication guides readers to the scratch, wear, and lubrication properties of polymers and the engineering applications, from biomedical research to automotive engineering. Author Hongbo Zeng details different experimental and theoretical methods used to probe static and dynamic properties of polymer materials and biomacromolecular systems. Topics include the use of atomic force microscopy (AFM) to analyze nanotribology, polymer thin films and brushes, nanoparticles, rubber and tire technology, synovial joint lubrication, adhesion in paper products, bioMEMS, and electrorheological fluid
Animated construction of line drawings
No full textRevealing the sketching sequence of a line drawing can be visually intriguing and used for video-based storytelling. Typically this is enabled based on tedious recording of artists' drawing process. We demonstrate that it is often possible to estimate a reasonable drawing order from a static line drawing with clearly defined shape geometry, which looks plausible to a human viewer. We map the key principles of drawing order from drawing cognition to computational procedures in our framework. Our system produces plausible animated constructions of input line drawings, with no or little user intervention. We test our algorithm on a range of input sketches, with varying degree of complexity and structure, and evaluate the results via a user study. We also present applications to gesture drawing synthesis and drawing animation creation especially in the context of video scribing. © 2011 ACM.We thank the reviewers for their constructive comments, the many participants of the user studies for their time, Gail Redman for video narration, Kanlun Cen for the line drawings used in this paper, and Wilmot Li for his many insightful comments. Hongbo Fu was supported by grants from CityU (No. SRG7002533), and the HKSAR Research Grants Council (No. 9041562); Ligang Liu was supported by the National Natural Science Foundation of China (61070071) and the 973 National Key Basic Research Foundation of China (2009CB320801); Niloy J. Mitra was partially supported by KAUST research network
Non-Local Low-Rank Normal Filtering for Mesh Denoising
No full textThis paper presents a non-local low-rank normal filtering method for mesh denoising. By exploring the geometric similarity between local surface patches on 3D meshes in the form of normal fields, we devise a low-rank recovery model that filters normal vectors by means of patch groups. In summary, our method has the following key contributions. First, we present the guided normal patch covariance descriptor to analyze the similarity between patches. Second, we pack normal vectors on similar patches into the normal-field patch-group (NPG) matrix for rank analysis. Third, we formulate mesh denoising as a low-rank matrix recovery problem based on the prior that the rank of the NPG matrix is high for raw meshes with noise, but can be significantly reduced for denoised meshes, whose normal vectors across similar patches should be more strongly correlated. Furthermore, we devise an objective function based on an improved truncated 'gamma' norm, and derive an optimization procedure using the alternative direction method of multipliers and iteratively re-weighted least squares techniques.We conducted several experiments to evaluate our method using various 3D models, and compared our results against several state-of-the-art methods. Experimental results show that our method consistently outperforms other methods and better preserves the fine details.Computer Graphics ForumMesh Denoising37
- …
