Texture Model Research Articles

Robust cross-vendor mammographic texture models using augmentation-based domain adaptation for long-term breast cancer risk.

Risk-stratified breast cancer screening might improve early detection and efficiency without comprising quality. However, modern mammography-based risk models do not ensure adaptation across vendor-domains and rely on cancer precursors, associated with short-term risk, which might limit long-term risk assessment. We report a cross-vendor mammographic texture model for long-term risk. The texture model was robustly trained using two systematically designed case-control datasets. Textural features, indicative of future breast cancer, were learned by excluding samples with diagnosed/potential malignancies from training. An augmentation-based domain adaption technique, based on flavorization of mammographic views, ensured generalization across vendor-domains. The model was validated in 66,607 consecutively screened Danish women with flavorized Siemens views and 25,706 Dutch women with Hologic-processed views. Performances were evaluated for interval cancers (IC) within 2 years from screening and long-term cancers (LTC) from 2 years after screening. The texture model was combined with established risk factors to flag 10% of women with the highest risk. In Danish women, the texture model achieved an area under the receiver operating characteristic curve (AUC) of 0.71 and 0.65 for ICs and LTCs, respectively. In Dutch women with Hologic-processed views, the AUCs were not different from AUCs in Danish women with flavorized views. The AUC for texture combined with established risk factors increased to 0.68 for LTCs. The 10% of women flagged as high-risk accounted for 25.5% of ICs and 24.8% of LTCs. The texture model robustly estimated long-term breast cancer risk while adapting to an unseen processed vendor-domain and identified a clinically relevant high-risk subgroup.

Texture representation via analysis and synthesis with generative adversarial networks

We investigate data-driven texture modeling via analysis and synthesis with generative adversarial networks. For network training and testing, we have compiled a diverse set of spatially homogeneous textures, ranging from stochastic to regular. We adopt StyleGAN3 for synthesis and demonstrate that it produces diverse textures beyond those represented in the training data. For texture analysis, we propose GAN inversion using a novel latent domain reconstruction consistency criterion for synthesized textures, and iterative refinement with a sliced Wasserstein loss for real-world textures. We propose perceptual procedures for evaluating network capabilities, exploring the global and local behavior of latent space trajectories, and comparing with existing texture analysis-synthesis techniques.

Effect of spherical crown texturing and ionized sulfur infiltration on the tribological performance of piston-cylinder liner

In this study, the piston-cylinder liner was treated with single glow photoionization sulfurization of a cast aluminum alloy (AlSi10Mg), and the effect of the spherical crown texture on frictional properties of the piston-cylinder liner were investigated before and after the sulfurization treatment. The treated specimens were characterized and evaluated via microscopic morphological analysis, three-dimensional morphological analysis, X-ray diffraction (XRD), thin film diffraction, microhardness testing, and friction wear experiments. Fluent fluid analysis software was used to develop a flow field simulation model for the spherical crown texture. Results of the study showed that the average friction coefficients of the specimens treated with sulfur infiltration were all lower than those of untreated specimens, and the low friction coefficients were closely related to the laminar structure and weak interfacial bonding of the sulfur infiltration layer. The spherical crown structure effectively affected the stree distribution on surface of the specimens uniformly, which improved the bearing capacity and durability of the specimens, and the surface texture helped reduce the contact area and store the debris generated during the wear process, thus improving the friction performance. Mechanical stamping caused material aggregation on the edges of the weave, which increased the hardness of the edges and thus reduces the increase in wear around the weave. The surface microstructure of the specimens was effectively improved after sulfurization treatment, leading to higher wear resistance for the sulfurized specimens. Given the relative lack of research on the composite ionic sulfurization of piston-cylinder liner spherical crown texture at home and abroad, this experiment provides data support for this direction. The results provide beneficial clues for the design and optimization of the piston-cylinder liner spherical crown structure.

Deep learning based multi-view stereo matching and 3D scene reconstruction from oblique aerial images

In this paper, we propose a practical three-dimensional (3D) real-scene reconstruction framework named Deep3D, which is paired with a deep learning based multi-view stereo (MVS) matching model named the adaptive multi-view aggregation matching (Ada-MVS) model, to obtain a 3D textured mesh model from multi-view oblique aerial images. Deep3D is the first deep learning based framework for 3D scene reconstruction, in which aerial triangulation and view selection are first performed on the input images, and the depth map of each image is then inferred using the pretrained Ada-MVS model. All the inferred depth maps are then fused into a dense point cloud after filtering the outliers. Finally, the 3D textured mesh is extracted from the dense 3D points as the final product. In the Ada-MVS model, a novel adaptive inter-view aggregation module is specially proposed to address the inconsistent information among oblique views and to fuse the multi-view costs into a robust cost volume. A lightweight recurrent regularization module is also designed for high-efficiency processing of high-capacity aerial images with large depth variations. Moreover, as oblique aerial image datasets are currently lacking, we built a large-scale synthetic multi-view oblique aerial image dataset (WHU-OMVS dataset) for deep learning based model training and methodology evaluation for the task of 3D scene reconstruction. The experimental results show that, firstly, the proposed Ada-MVS model has obvious advantages when used with high-capacity oblique aerial images, compared with several relevant learning-based MVS methods. Secondly, through a comprehensive comparison with popular commercial software packages and open-source solutions, it is shown that the proposed Deep3D framework outperforms all the other solutions in terms of reconstruction quality, and outperforms all the open-source solutions and some of the software packages in terms of efficiency on the WHU-OMVS dataset. Thirdly, the Deep3D framework shows a stable generalization ability and excellent performance when applied to other oblique or nadir aerial images, without any further fine-tuning. The dataset and code will be available at http://gpcv.whu.edu.cn/data.

Effect of Groove Structure on Lubrication Performance of Water-Lubricated Stern Tube Bearings

This study investigated the lubrication characteristics (i.e., the groove ratio and width) of water-lubricated stern tube bearings, based on the flexibility matrix method and lubrication theory. Considering the elastic deformation of the lining, a fluid structure interaction (FSI) model of the surface micro-groove texture of a water-lubricated stern tube bearing was established, and the correctness and rationality of the model were verified by experiments. Micro-grooved, surface-lubricated, water-lubricated stern tube bearings with three different cross-sectional shapes (rectangular, circular, and triangular) were designed. The influences of the groove area ratio and width on the bearing load-carrying capacity and friction coefficient were analyzed. At a groove area ratio of 0.31, the load-carrying capacity of the rectangular grooved stern tube bearing reached the maximum value and the friction coefficient reached the minimum value. It is recommended to design and use water-lubricated stern tube bearings, especially Thordon water-lubricated stern tube bearings, with rectangular micro-grooves, with a groove area ratio of 0.30–0.32, so that the best lubrication performance can be obtained. With the increase in the micro-groove width, the lubrication of water-lubricated stern tube bearings with partial rectangular micro-grooves is significantly better than that of others. Under the same conditions, the bearing load-carrying capacity and friction performance of local groove stern bearings is significantly better than that of global groove stern bearings.

Image compression and reconstruction based on GUI and Huffman coding

Huffman coding is an important part of image compression technology, the image compression platform is based on GUI, and Huffman is also widely used. This paper introduces the basic principle of the Huffman algorithm, compares it with arithmetic coding and run length encoding, and expounds on the application of these three algorithms in JPEG compression. The AC algorithm combined block-based, fine texture models and adaptive arithmetic coding in the given an example. The RLE algorithm used automatic threshold, direction judgment, and selective value counts to improve its compression efficiency. JPEG algorithm adopted an adaptive quantization table to reduce the distortion rate of image compression. This paper proves the possibility of a better compression rate and distortion rate of the hybrid compression algorithm by demonstrating the improved example of the basic image compression algorithm. In the future, the improved basic algorithms can be combined based on the original JPEG algorithm, and different algorithms can be integrated on the GUI to face different use environments.

Estimation and mapping of soil texture content based on unmanned aerial vehicle hyperspectral imaging

Soil texture is one of the important physical and natural properties of soil. Much of the current research focuses on soil texture monitoring using non-imaging geophysical spectrometers. However there are fewer studies utilizing unmanned aerial vehicle (UAV) hyperspectral data for soil texture monitoring. UAV mounted hyperspectral cameras can be used for quickly and accurately obtaining high-resolution spatial information of soil texture. A foundation has been laid for the realization of rapid soil texture surveys using unmanned airborne hyperspectral data without field sampling. This study selected three typical farmland areas in Huangshui Basin of Qinghai as the study area, and a total of 296 soil samples were collected. Data calibration of UAV spectra using laboratory spectra and field in situ spectra to explore the feasibility of applying laboratory soil texture models directly to field conditions. This results show that UAV hyperspectral imagery combined with machine learning can obtain a set of ideal processing methods. The pre-processing of the spectral data can obtain high accuracy of soil texture estimation and good mapping effect. The results of this study can provide effective technical support and decision-making assistance for future agricultural land planning on the Tibetan Plateau. The main innovation of this study is to establish a set of processing procedures and methods applicable to UAV hyperspectral imagery to provide data reference for monitoring soil texture in agricultural fields on the Tibetan Plateau.

Model of Surface Texture for Honed Gear Considering Motion Path and Geometrical Shape of Abrasive Particle

Gear power-honing is mainly applied to finish small and medium-sized automotive gears, especially in new energy vehicles. The distinctive curved surface texture greatly improves the noise emission and service life of honed gears. The surface texture for honed gear considering the motion path and geometrical shape of abrasive particles has not been investigated. In this paper, the kinematics of the gear honing process is analyzed, and the machining marks produced by the abrasive particles of honing wheel scratching abrasive particles against the workpiece gear are calculated. The tooth surface roughness is modeled considering abrasive particle shapes and material plastic pile-ups. This results in a mathematical model that characterizes the structure of the tooth surface and the orientation of the machining marks. Experiments were used to verify the model, with a maximum relative error of less than 10% when abrasive particles are spherical. Based on this model, the effects of process parameters on the speeds of discrete points on the tooth flank, orientations of machining marks and roughness are discussed. The results show that the shaft angle between the workpiece gear and the honing wheel and the speed of the honing wheel is the main process parameters affecting the surface texture. This research proposes a surface texture model for honed gear, which can provide a theoretical basis for optimizing process parameters for gear power-honing.

Geometrical and crystal plasticity modelling: Towards the establishment of a process-structure-property relationship for additively manufactured 316L struts

This paper presents a methodology to establish a process structure property (PSP) relationship for the additive manufacturing (AM) of thin AISI 316L struts, as might be used in coronary stent applications. The methodology is based on a new geometrically based process-structure method for AM process variables, and crystal plasticity finite element (CPFE) modelling that includes the representation of melt pool microstructure morphology and texture. The effects of AM process variables are characterised with respect to ductility, yield strength and UTS across a range of process variables, including hatch spacing, layer thickness and build orientation (two orientations are considered: horizontal and vertical). The CPFE-predicted effect of build orientation is shown to be consistent with experimental test results. The present methodology has allowed identification of optimal hatch spacing and layer thickness for a given laser spot size, power and scanning speed. CPFE modelling of AM melt pool texture is shown to be key to successful prediction of the effects of process variables on mechanical behaviour.

Estimating the vertical direction in a photogrammetric 3D model, with application to visualization

The process of building a Virtual Reality (VR) environment from images involves several steps: choose experimental conditions (scene, camera, trajectory, weather), take the images, reconstruct a textured 3D model thanks to a photogrammetry software, and import the 3D model into a game engine. This paper focuses on a postprocessing technique for the photogrammetry step, mostly for outdoor environments that cannot be reconstructed using an unmanned aerial vehicle. As visualization applications (including VR) need a 3D model with a known vertical direction, a method is introduced to compute it. The method is based on 3D principal component analysis and a 2D Hough transform. In the experiments, we first reconstruct both man-made and natural immersive environments using a helmet-held 360 camera, then we import the 3D models in good coordinate systems (i.e. with a vertical axis and a plausible scale) into Unity, and finally we use VR headsets to explore the scenes like a pedestrian. We also experiment on scanner data and show that our method is competitive with previous work.

High-Resolution Volumetric Reconstruction for Clothed Humans

We present a novel method for reconstructing clothed humans from a sparse set of, e.g., 1–6 RGB images. Despite impressive results from recent works employing deep implicit representation, we revisit the volumetric approach and demonstrate that better performance can be achieved with proper system design. The volumetric representation offers significant advantages in leveraging 3D spatial context through 3D convolutions, and the notorious quantization error is largely negligible with a reasonably large yet affordable volume resolution, e.g., 512. To handle memory and computation costs, we propose a sophisticated coarse-to-fine strategy with voxel culling and subspace sparse convolution. Our method starts with a discretized visual hull to compute a coarse shape and then focuses on a narrow band nearby the coarse shape for refinement. Once the shape is reconstructed, we adopt an image-based rendering approach, which computes the colors of surface points by blending input images with learned weights. Extensive experimental results show that our method significantly reduces the mean point-to-surface (P2S) precision of state-of-the-art methods by more than 50% to achieve approximately 2mm accuracy with a 512 volume resolution. Additionally, images rendered from our textured model achieve a higher peak signal-to-noise ratio (PSNR) compared to state-of-the-art methods.

Synergistic Use of Sentinel-1 and Sentinel-2 Based on Different Preprocessing for Predicting Forest Aboveground Biomass

Forest aboveground biomass (AGB, Mg/ha) measurement is one of the key indicators for carbon storage evaluation. Remote sensing techniques have been widely employed to predict forest AGB. However, little attention has been paid to the implications involved in the preprocessing of satellite data. In this work, considering the areas of low forest AGB in our survey plots, we explored the implications of employing atmospheric correction and speckle filtering with Sentinel-1 (S1) synthetic aperture radar (SAR) and Sentinel-2 (S2) to predict forest AGB using multiple linear regression (MLR) and extreme gradient boosting (XGBoost). In the present study, the types of plots examined included oaks (Quercus spp.), Chinese firs (Cunninghamia lanceolata), and Masson pines (Pinus massoniana), and all of the plots were investigated. Specifically, the feature variables related to S1 (dual polarization and texture measures) and S2 (spectral bands) were modeled individually, and 16 feature sets, including different combinations of S1 and S2 based on different preprocessing measures, were established using MLR and XGBoost. The results show that speckle filtering and atmospheric correction marginally influenced the capacity of the S2 spectral bands, the SAR dual-polarization backscatter, and the SAR-based textural measures in predicting the AGB in our survey plots. The associations between the speckle-filtered and unfiltered SAR images and the S2 Top-of-Atmosphere and Bottom-of-Atmosphere products were considerably strong. Additionally, the texture models generally showed better performances than the raw SAR data. Ultimately, the groups that only encompassed the S2 spectral bands were the best-performing groups among the 16 feature sets, while the groups that included only S1-based data generally performed the worst.

Assessing Breast Cancer Risk by Combining AI for Lesion Detection and Mammographic Texture.

Background Recent mammography-based risk models can estimate short-term or long-term breast cancer risk, but whether risk assessment may improve by combining these models has not been evaluated. Purpose To determine whether breast cancer risk assessment improves when combining a diagnostic artificial intelligence (AI) system for lesion detection and a mammographic texture model. Materials and Methods This retrospective study included Danish women consecutively screened for breast cancer at mammography from November 2012 to December 2015 who had at least 5 years of follow-up data. Examinations were evaluated for short-term risk using a commercially available diagnostic AI system for lesion detection, which produced a score to indicate the probability of cancer. A mammographic texture model, trained on a separate data set, assessed textures associated with long-term cancer risk. Area under the receiver operating characteristic curve (AUC) analysis was used to evaluate both the individual and combined performance of the AI and texture models for the prediction of future cancers in women with a negative screening mammogram, including those with interval cancers diagnosed within 2 years of screening and long-term cancers diagnosed 2 years or more after screening. AUCs were compared using the DeLong test. Results The Danish screening cohort included 119 650 women (median age, 59 years [IQR, 53-64 years]), of whom 320 developed interval cancers and 1401 developed long-term cancers. The combination model achieved a higher AUC for interval and long-term cancers grouped together than either the diagnostic AI (AUC, 0.73 vs 0.70; P < .001) or the texture risk (AUC, 0.73 vs 0.66; P < .001) models. The 10% of women with the highest combined risk identified by the combination model accounted for 44.1% (141 of 320) of interval cancers and 33.7% (472 of 1401) of long-term cancers. Conclusion Combining a diagnostic AI system and mammographic texture model resulted in improved risk assessment for interval cancers and long-term cancers and enabled identification of women at high risk. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Poynton and Slanetz in this issue.

MILK: a Python scripting interface to MAUD for automation of Rietveld analysis.

Modern diffraction experiments (e.g. in situ parametric studies) present scientists with many diffraction patterns to analyze. Interactive analyses via graphical user interfaces tend to slow down obtaining quantitative results such as lattice parameters and phase fractions. Furthermore, Rietveld refinement strategies (i.e. the parameter turn-on-off sequences) tend to be instrument specific or even specific to a given dataset, such that selection of strategies can become a bottleneck for efficient data analysis. Managing multi-histogram datasets such as from multi-bank neutron diffractometers or caked 2D synchrotron data presents additional challenges due to the large number of histogram-specific parameters. To overcome these challenges in the Rietveld software Material Analysis Using Diffraction (MAUD), the MAUD Interface Language Kit (MILK) is developed along with an updated text batch interface for MAUD. The open-source software MILK is computer-platform independent and is packaged as a Python library that interfaces with MAUD. Using MILK, model selection (e.g. various texture or peak-broadening models), Rietveld parameter manipulation and distributed parallel batch computing can be performed through a high-level Python interface. A high-level interface enables analysis workflows to be easily programmed, shared and applied to large datasets, and external tools to be integrated with MAUD. Through modification to the MAUD batch interface, plot and data exports have been improved. The resulting hierarchical folders from Rietveld refinements with MILK are compatible with Cinema: Debye-Scherrer, a tool for visualizing and inspecting the results of multi-parameter analyses of large quantities of diffraction data. In this manuscript, the combined Python scripting and visualization capability of MILK is demonstrated with a quantitative texture and phase analysis of data collected at the HIPPO neutron diffractometer.

Piglet Model for Modified Blalock-Taussig-Thomas Shunt Training

BackgroundThe Blalock-Taussig-Thomas shunt is one of the milestone procedures in cardiac surgery and a crucial operation for palliating blue babies. It is currently impractical to train new surgeons on infants even under the supervision of their mentors. A realistic training model for location, exposure, tissue toughness, and texture is required. The objective of this study was to determine whether the model is realistic and practical and to evaluate the quality of anastomosis performed by surgical trainees. MethodsPiglet heart and lung models were selected from piglets of 3 to 5 kg. Six cardiothoracic surgical residents participated in the program and performed the procedure. Subsequently, every trainee answered a standard questionnaire in terms of satisfaction in reality and practicality; the trainers evaluated the quality of the anastomosis. The median score of each item was analyzed and described. The quality of anastomosis was evaluated and analyzed similarly. ResultsMedian scores of satisfaction (14 questions) were 9, 10, 9, 9.5, 10, 9.5, 10, 9.5, 10, 8.5, 10, 8.5, 10, and 8. Median challenge scores (6 questions) were 10, 10, 10, 10, 10, 9.5, and 10. The median anastomosis evaluation score from 6 trainees was 25 (24-25). ConclusionsOur heart-lung simulation model is perhaps 1 of the most realistic simulations in practicality, visualization, exposure of the related structures, similar scenes to the real operation, and sense of touch and texture. Realistic models should be developed, but they should be inexpensive.

Photogrammetry for 3D Reconstruction of Objects: Effects of Geometry, Texture and Photographing

Today, the use of 3D scanning methods is increasing in a wide range of industries from biomedical to game production. The photogrammetry method is in demand by 3D scanning users due to its portability, photorealistic textured modeling, and low cost. Scanned objects vary according to industries and can have various materials, geometries and textures. In order to better benefit from the advantages of the photogrammetry method, it is necessary to know the response of the method against different parameters. Unlike the studies in the literature, in this study, many objects with various geometries, different materials and textures were reconstructed and evaluated by using the same equipment and workflow in the same environment, and thus a fair evaluation was performed for various factors affecting the reconstruction process. In this context, 24 differently shaped objects with metal, fabric, plastic, wood, glass, ceramic and organic textures were reconstructed. A comprehensive evaluation was conducted on the solid and textured models generated from the objects. In addition to the effects of object geometry, texture properties, photo shooting angle and distance, it also sought answers to questions about factors affecting the accuracy and mesh quality of the 3D models. In this study, it is possible to find some recent specific results and validated general results together related to photogrammetric modeling.

Modeling of dynamic recrystallization texture in hot extruded Mg

Because of the strong effect of texture on the properties of alloys, the evolution of texture during hot extrusion of Mg has been studied through experiments and simulations. For a better understanding of the texture evolution at high processing temperatures, the model based on viscoplastic formulation was extended to a cellular automaton procedure to consider the recrystallization mechanisms. Texture evolution was simulated for the basal, prismatic and pyramidal 2nd order slip systems. Twinning was not considered because no twins were observed in the microstructural investigation. In the experimental investigations, the samples were extruded to a strain of ε = 2.7 at different temperatures, namely 180 °C, 280 °C, and 380 °C. The microstructure, texture, and hardness were evaluated by optical and electron microscopy using EBSD, X-ray diffraction, and Vickers tests, respectively. The main textural components were identified for each deformation temperature. Based on the simulations, a mechanism for the formation of a specific component was proposed. It was found that the change in the main texture component depends not only on the critical resolved shear stress for a specific slip system but also on the recrystallization parameters. The vpsc model has been shown to predict only the deformation texture components because the temperature is accounted for in accordance with the critical resolved shear stress and rate sensitivity. These two parameters affect the slip system activity. The new model using the cellular automaton procedure, which covers recrystallization processes, produces textures that fall in line with the experimental results.

Face Forgery Detection by 3D Decomposition and Composition Search.

Detecting digital face manipulation has attracted extensive attention due to fake media's potential risks to the public. However, recent advances have been able to reduce the forgery signals to a low magnitude. Decomposition, which reversibly decomposes an image into several constituent elements, is a promising way to highlight the hidden forgery details. In this paper, we investigate a novel 3D decomposition based method that considers a face image as the production of the interaction between 3D geometry and lighting environment. Specifically, we disentangle a face image into four graphics components including 3D shape, lighting, common texture, and identity texture, which are respectively constrained by 3D morphable model, harmonic reflectance illumination, and PCA texture model. Meanwhile, we build a fine-grained morphing network to predict 3D shapes with pixel-level accuracy to reduce the noise in the decomposed elements. Moreover, we propose a composition search strategy that enables an automatic construction of an architecture to mine forgery clues from forgery-relevant components. Extensive experiments validate that the decomposed components highlight forgery artifacts, and the searched architecture extracts discriminative forgery features. Thus, our method achieves the state-of-the-art performance.

CULTURAL HERITAGE DISSEMINATION: BIM MODELLING AND AR APPLICATION FOR A DIACHRONIC TALE

Abstract. The research purpose is to present a project of cultural dissemination and enhancement of the “Madonna della Pace” Sanctuary at Rocchetta di Airuno (Lecco - Italy), based on an immersive experience of knowledge of the history and places that characterize not only the sanctuary but also the “Cammino di Sant’Agostino” (of which it is one of the stages). The research goal focuses on the direct employment and exploitation of HBIM models for the digital fruition project. An integrated digital survey based on a terrestrial laser scanner and photogrammetry was conducted to provide a complete geometrical representation of the sanctuary and its surroundings. Both output point clouds were employed as metric and geometric references to create the reality-based parametric model. Specifically, the work focuses on creating a three-dimensional chronological model of the sanctuary, which not only represents the current state of the cultural asset but is also enriched through the definition of different evolutionary phases of the architectural artifact based on an indepth study of the photographic and bibliographic documentations. Four Project Phasing has been identified to represent the most significant transformations of the building and were managed using the time parameter in the same BIM project. The sanctuary geometric and parametric models were displayed and navigable thanks to the aid of Virtual and Augmented Reality applications. A VR environment was defined to display in the first person the textured model. Finally, an AR smartphone app prototype was developed to show tourists the sanctuary’s historical transformation over time.

DOVE: Learning Deformable 3D Objects by Watching Videos

Learning deformable 3D objects from 2D images is often an ill-posed problem. Existing methods rely on explicit supervision to establish multi-view correspondences, such as template shape models and keypoint annotations, which restricts their applicability on objects “in the wild”. A more natural way of establishing correspondences is by watching videos of objects moving around. In this paper, we present DOVE, a method that learns textured 3D models of deformable object categories from monocular videos available online, without keypoint, viewpoint or template shape supervision. By resolving symmetry-induced pose ambiguities and leveraging temporal correspondences in videos, the model automatically learns to factor out 3D shape, articulated pose and texture from each individual RGB frame, and is ready for single-image inference at test time. In the experiments, we show that existing methods fail to learn sensible 3D shapes without additional keypoint or template supervision, whereas our method produces temporally consistent 3D models, which can be animated and rendered from arbitrary viewpoints. Project page: https://dove3d.github.io/.