RealFusion: 360° Reconstruction of Any Object from a Single Image

Melas-Kyriazi, Luke; Rupprecht, Christian; Laina, Iro; Vedaldi, Andrea

doi:10.48550/arxiv.2302.10663

Cited by 3 publications

(9 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Diffusion models are a class of generative models that make use of a Markovian noising process to iteratively reverse the noise. In recent years, several researchers [28]- [31] have explored the use of diffusion models in conjunction with radiance fields and have demonstrated excellent results in tasks such as conditional synthesis, completion, and other related tasks [32], [33].…”

Section: B Novel View Synthesismentioning

confidence: 99%

Single-View 3D Garment Reconstruction Using Neural Volumetric Rendering

Chen,

Xie,

Yang

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Reconstructing 3D garment models usually requires laborious data-fetching processes, such as expensive lidar, multiple-view images, or SMPL models of the garments. In this paper, we propose a neat framework that takes single-image inputs for generating pseudo-sparse views of 3D garments and synthesizing multi-view images into a high-quality 3D neural model. Specifically, our framework combines a pretrained pseudo sparse view generator and a volumetric signed distance function (SDF) representationbased network for 3D garment modeling, which uses neural networks to represent both the density and radiance fields. We further introduce a stride fusion strategy to minimize the pixel-level loss in key viewpoints and semantic loss in random viewpoints, which produces view-consistent geometry and sharp texture details. Finally, a multi-view rendering module utilizes the learned SDF representation to generate multi-view garment images and extract accurate mesh and texture from them. We evaluate our proposed framework on the Deep Fashion 3D dataset and achieve state-of-the-art performance in terms of both quantitative and qualitative evaluations.

show abstract

Section: B Novel View Synthesismentioning

confidence: 99%

Single-View 3D Garment Reconstruction Using Neural Volumetric Rendering

Chen,

Xie,

Yang

et al. 2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…The most related to our work is DreamFusion [26], which introduced Score Distillation Sampling (SDS) for creation of 3D assets, leveraging the power of text-to-image diffusion models. Despite the flexible merit of SDS to enable the optimization of arbitrary differentiable operators, most works mainly focus on applying SDS to enhance the synthesis quality of 3D scenes by introducing 3D specific frameworks [48,49,50,51,52]. Although there exists some work to apply SDS for visual domains other than 3D scenes, they have limited their scope to image editing [53], or image generation [54].…”

Section: Related Workmentioning

confidence: 99%

Covert Brain Infarction as a Risk Factor for Stroke Recurrence in Patients With Atrial Fibrillation

Kim

Han

Jeong

et al. 2023

Stroke

View full text Add to dashboard Cite

Background: We aimed to evaluate covert brain infarction (CBI), frequently encountered during the diagnostic work-up of acute ischemic stroke, as a risk factor for stroke recurrence in patients with atrial fibrillation (AF). Methods: For this prospective cohort study, from patients with acute ischemic stroke hospitalized at 14 centers between 2017 and 2019, we enrolled AF patients without history of stroke or transient ischemic attack and divided them into the CBI (+) and CBI (−) groups. The 2 groups were compared regarding the 1-year cumulative incidence of recurrent ischemic stroke and all-cause mortality using the Fine and Gray subdistribution hazard model with nonstroke death as a competing risk and the Cox frailty model, respectively. Each CBI lesion was also categorized into either embolic-appearing (EA) or non-EA pattern CBI. Adjusted hazard ratios and 95% CIs of any CBI, EA pattern CBI only, non-EA pattern CBI only, and both CBIs were estimated. Results: Among 1383 first-ever stroke patients with AF, 578 patients (41.8%) had CBI. Of these 578 with CBI, EA pattern CBI only, non-EA pattern CBI only, and both CBIs were 61.8% (n=357), 21.8% (n=126), and 16.4% (n=95), respectively. The estimated 1-year cumulative incidence of recurrent ischemic stroke was 5.2% and 1.9% in the CBI (+) and CBI (−) groups, respectively ( P =0.001 by Gray test). CBI increased the risk of recurrent ischemic stroke (adjusted hazard ratio [95% CI], 2.91 [1.44–5.88]) but did not the risk of all-cause mortality (1.32 [0.97–1.80]). The EA pattern CBI only and both CBIs elevated the risk of recurrent ischemic stroke (2.76 [1.32–5.77] and 5.39 [2.25–12.91], respectively), while the non-EA pattern only did not (1.44 [0.40–5.16]). Conclusions: Our study suggests that AF patients with CBI might have increased risk of recurrent stroke. CBI could be considered when estimating the stroke risk in patients with AF.

show abstract

“…To leverage the capability of 2D generative models, one line of approaches [Melas-Kyriazi et al 2023;Poole et al 2022; propose Distillation Sampling to generate 3D assets from texts. These approaches usually suffer from low diversity, oversaturation, and multi-face problems.…”

Section: Multi-view/3d Generation With 2d Generationmentioning

confidence: 99%

“…Diffusion models are a class of generative models that make use of a Markovian noising process to iteratively reverse the noise. In recent years, several researchers Lin et al 2023b;Liu et al 2023b;Melas-Kyriazi et al 2023;Poole et al 2022;Shi et al 2023a] have explored the use of diffusion models in conjunction with radiance fields and have demonstrated excellent results in tasks such as conditional synthesis, completion, and other related tasks [Zeng et al 2022;Zhou et al 2021].…”

Section: Novel View Synthesismentioning

confidence: 99%

“…Recently, there has been remarkable progress in generating 3D objects thanks to the development of generation models like diffusion and GAN [Ho et al 2020;Karras et al 2021Karras et al , 2019Ramesh et al 2022;Rombach et al 2022]. These promising results soon attracted a lot of attention and led to many intriguing possibilities, such as generating 3D objects from a single image, text prompts, or environment Melas-Kyriazi et al 2023;Poole et al 2022;]. Many of these 3D generation works leverage 2D image generation models; they usually consist of two stages: an image generation stage and a 3D reconstruction stage, and face one major challenge of reconstructing 3D objects from the generated 2D images, i.e.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Eye diseases burden in China in the past 30 years

Chen¹,

Lou²,

Ye³

2021

J Zhejiang Univ (Med Sci)

View full text Add to dashboard Cite

Objective: To explore the national burden of eye diseases in China and its trends from 1990 to 2019. Methods: The national burden of eye diseases in China, including case numbers, prevalence rate, age-standardized prevalence rate (ASR), disability-adjusted life years (DALY), DALY rate and age-standardized DALY rate (ASD) were calculated and stratified by sex and age. The trends of eye diseases burden from 1990 to 2019 and the correlation between eye diseases burden and human development index (HDI) were analyzed. Results: In 2019, the total case number of eye diseases in China was 0.21 billion, the ASR was 9511/10 5 , the total number of DALY was

show abstract

RealFusion: 360° Reconstruction of Any Object from a Single Image

Cited by 3 publications

References 38 publications

Single-View 3D Garment Reconstruction Using Neural Volumetric Rendering

Single-View 3D Garment Reconstruction Using Neural Volumetric Rendering

Covert Brain Infarction as a Risk Factor for Stroke Recurrence in Patients With Atrial Fibrillation

Eye diseases burden in China in the past 30 years

Contact Info

Product

Resources

About