Search for: All records

We present WonderWorld, a novel framework for interactive 3D scene generation that enables users to interactively specify scene contents and layout and see the created scenes in low latency. The major challenge lies in achieving fast generation of 3D scenes. Existing scene generation approaches fall short of speed as they often require (1) progressively generating many views and depth maps, and (2) time-consuming optimization of the scene representations. Our approach does not need multiple views, and it leverages a geometry-based initialization that significantly reduces optimization time. Another challenge is generating coherent geometry that allows all scenes to be connected. We introduce the guided depth diffusion that allows partial conditioning of depth estimation. WonderWorld generates connected and diverse 3D scenes in less than 10 seconds on a single A6000 GPU, enabling real-time user interaction and exploration. Our interactive demo, full code, data, and software can be found at https://kovenyu.com/WonderWorld/ 

Realistic object interactions are crucial for creating immersive virtual experiences, yet synthesizing realistic 3D object dynamics in response to novel interactions remains a significant challenge. Unlike unconditional or text-conditioned dynamics generation, action-conditioned dynamics requires perceiving the physical material properties of objects and grounding the 3D motion prediction on these properties, such as object stiffness. However, estimating physical material properties is an open problem due to the lack of material ground-truth data, as measuring these properties for real objects is highly difficult. We present PhysDreamer, a physics-based approach that endows static 3D objects with interactive dynamics by leveraging the object dynamics priors learned by video generation models. By distilling these priors, PhysDreamer enables the synthesis of realistic object responses to novel interactions, such as external forces or agent manipulations. We demonstrate our approach on diverse examples of elastic objects and evaluate the realism of the synthesized interactions through a user study. PhysDreamer takes a step towards more engaging and realistic virtual experiences by enabling static 3D objects to dynamically respond to interactive stimuli in a physically plausible manner. See our project page at this https URL. 

Oxidation of the sub-arc mantle driven by slab-derived fluids has been hypothesized to contribute to the formation of gold deposits in magmatic arc environments that host the majority of metal resources on Earth. However, the mechanism by which the infiltration of slab-derived fluids into the mantle wedge changes its oxidation state and affects Au enrichment remains poorly understood. Here, we present the results of a numerical model that demonstrates that slab-derived fluids introduce large amounts of sulfate (S6+) into the overlying mantle wedge that increase its oxygen fugacity by up to 3 to 4 log units relative to the pristine mantle. Our model predicts that as much as 1 wt.% of the total dissolved sulfur in slab-derived fluids reacting with mantle rocks is present as the trisulfur radical ion, S3–. This sulfur ligand stabilizes the aqueous Au(HS)S3– complex, which can transport Au concentrations of several grams per cubic meter of fluid. Such concentrations are more than three orders of magnitude higher than the average abundance of Au in the mantle. Our data thus demonstrate that an aqueous fluid phase can extract 10 to 100 times more Au than in a fluid-absent rock-melt system during mantle partial melting at redox conditions close to the sulfide-sulfate boundary. We conclude that oxidation by slab-derived fluids is the primary cause of Au mobility and enrichment in the mantle wedge and that aqueous fluid-assisted mantle melting is a prerequisite for formation of Au-rich magmatic hydrothermal and orogenic gold systems in subduction zone settings. 

Abstract In June 2020, the tropical Atlantic and the Caribbean Basin were affected by a series of African dust outbreaks unprecedented in size and intensity. These events, informally named “Godzilla”, coincided with CALIMA, a large field campaign, offering a rare opportunity to assess the impact of African dust on air quality in the Greater Caribbean Basin. Network measurements of respirable particles (i.e., PM₁₀and PM_2.5) showed that dust significantly degraded regional air quality and increased the risk to public health in the Caribbean, the southern United States, northern South America, and Central America. CALIMA examined the meteorological context of Godzilla dust events over North Africa and how these conditions might relate to the greatly increased dust emissions and enhanced transport to the Americas. Godzilla was linked to strong pressure anomalies over West Africa, resulting in a large-scale geostrophic wind anomaly at 700 hPa over North Africa. We used surface-based and columnar measurements to test the performance of two frequently used aerosol forecast models: the NASA GEOS and WRF-Chem models. The models showed some skills, but differed substantially between their forecasts, suggesting large uncertainties in these forecasts that are critical for issuing early warnings of health-threatening dust events. Our results demonstrate the value of an integrated approach in characterizing the spatial and temporal variability of African dust transport and assessing its impact on regional air quality. Future studies are needed to improve models and to track the long-term changes in dust transport from Africa under a changing climate. 

Dynamic spectrum sharing has emerged as a promising solution to address the spectrum scarcity challenge. Currently, the FCC has designated several Spectrum Access Systems (SAS) administrators to deploy their SAS that coordinates the usage of the certificated shared band(s) such as the 3.55-3.7 GHz CBRS band. The SAS ensures that the incumbent’s access to the shared band is guaranteed while also granting commercial users access rights when the incumbents are not present. However, explicitly sharing the spectrum band(s) information among participants raises privacy concerns. Certain participants, such as curious SAS administrators, have the ability to deduce the confidential operational patterns of the incumbents through the Environmental Sensing Capability (ESC) or Incumbent Informing Capability (IIC) notifications. Additionally, a curious SAS administrator may obtain the client’s operational information of other SAS administrators throughout the process of inter-SAS coordination. We propose Pri-Share, a novel privacy-preserving spectrum sharing paradigm that tailors the threshold-based private set union (PSU) and homomorphic encryption (HE) techniques to address the aforementioned privacy problems. Specifically, it enables all parties to jointly compute a unified spectrum allocation plan to resolve the potential conflicts between different parties while safeguarding the confidentiality of each stakeholder’s spectrum requirements and usage. Pri-Share also ensures that while a curious participant might ascertain the usage of a particular spectrum band, they are unable to deduce the precise identity of the party utilizing it. Besides, Pri-Share adheres to the key spectrum allocation regulations outlined by FCC (part 96), such as assurance of access rights for various priority levels. Our implementation result shows that Pri-Share can be achieved with notable computational and communication efficiency, 

Light Field Networks, the re-formulations of radiance fields to oriented rays, are magnitudes faster than their coordinate network counterparts, and provide higher fidelity with respect to representing 3D structures from 2D observations. They would be well suited for generic scene representation and manipulation, but suffer from one problem: they are limited to holistic and static scenes. In this paper, we propose the Dynamic Light Field Network (DyLiN) method that can handle non-rigid deformations, including topological changes. We learn a deformation field from input rays to canonical rays, and lift them into a higher dimensional space to handle discontinuities. We further introduce CoDyLiN, which augments DyLiN with controllable attribute inputs. We train both models via knowledge distillation from pretrained dynamic radiance fields. We evaluated DyLiN using both synthetic and real world datasets that include various non-rigid deformations. DyLiN qualitatively outperformed and quantitatively matched state-of-the-art methods in terms of visual fidelity, while being 25 - 71x computationally faster. We also tested CoDyLiN on attribute annotated data and it surpassed its teacher model. Project page: https://dylin2023.github.io.