Single-view depth estimation from omnidirectional images has gained popularity with its wide range of applications such as autonomous driving and scene reconstruction. Although data-driven learning-based methods demon...
详细信息
ISBN:
(纸本)9781665496179
Single-view depth estimation from omnidirectional images has gained popularity with its wide range of applications such as autonomous driving and scene reconstruction. Although data-driven learning-based methods demonstrate significant potential in this field, scarce training data and ineffective 360 estimation algorithms are still two key limitations hindering accurate estimation across diverse domains. In this work, we first establish a large-scale dataset with varied settings called Depth360 to tackle the training data problem. This is achieved by exploring the use of a plenteous source of data, 360 videos from the internet, using a test-time training method that leverages unique information in each omnidirectional sequence. With novel geometric and temporal constraints, our method generates consistent and convincing depth samples to facilitate single-view estimation. We then propose an end-to-end two-branch multi-task learning network, SegFuse, that mimics the human eye to effectively learn from the dataset and estimate high-quality depth maps from diverse monocular RGB images. With a peripheral branch that uses equirectangular projection for depth estimation and a foveal branch that uses cubemap projection for semantic segmentation, our method predicts consistent global depth while maintaining sharp details at local regions. Experimental results show favorable performance against the state-of-the-art methods.
The goal of intrinsic image decomposition is to recover low level features of images. Most of the studies tend to consider only reflectance and shading, even though it is known that increasing the number of intrinsics...
详细信息
ISBN:
(数字)9781665496209
ISBN:
(纸本)9781665496209
The goal of intrinsic image decomposition is to recover low level features of images. Most of the studies tend to consider only reflectance and shading, even though it is known that increasing the number of intrinsics is beneficial for many applications. Existing intrinsic image datasets are quite limited. In this study, a dataset is introduced to provide a comprehensive benchmark to the field of intrinsic image decomposition. IID-NORD contains a large number of scenes and for each scene ground truth reflectance, shading, surface normal vectors, light vectors and depth map is provided to allow detailed decomposition. Moreover, diverse illuminants, viewing angles, and dynamic shadows are used to prevent any bias. To the best of available knowledge, IID-NORD is the most comprehensive dataset in the field of intrinsic image decomposition. IID-NORD will be available on the first author's official webpage.
When a 3D scene is captured in several 2D images, a compact description (or parameters) of the 3D scene can be estimated from the images. Such an inference is formulated as the inverse of rendering computer graphics a...
详细信息
ISBN:
(纸本)9783031063817;9783031063800
When a 3D scene is captured in several 2D images, a compact description (or parameters) of the 3D scene can be estimated from the images. Such an inference is formulated as the inverse of rendering computer graphics and is important for various applications, such as object recognition, inspection, and/or VR. In the present paper, we extend a photometric stereo method in such a way as to estimate the texture of the object in addition to previous estimation of parameters describing the objects and light sources. To do so, we need a realistic minimization method, combined with a method to obtain the Jacobian of the cost function with respect to the texture. We implemented this method and verified the validity of the framework using synthetic and real-world data.
The de facto standard for storing human motion data on a computer involves a representation based on Euler angles. This representation, while effective, has several shortcomings. Triplets of Euler angles are not uniqu...
详细信息
ISBN:
(数字)9788396242396
ISBN:
(纸本)9788396242396
The de facto standard for storing human motion data on a computer involves a representation based on Euler angles. This representation, while effective, has several shortcomings. Triplets of Euler angles are not unique, and the same posture may be expressed using different combinations of angles. Furthermore, many possible Euler angle triplets correspond to unnatural positions for human joints. This means that, in general, a large part of the representational space remains unused. In this paper, we further investigate a recently proposed representation inspired by molecular representations. It uses only two (instead of three) degrees of freedom per joint: a vector and a torsion angle. Using the two key ingredients of this new representation, we present a complete analysis of the graphics Lab Motion Capture Database. The data found in this analysis provide us with some powerful insights about natural and unnatural human postures in human motions. These insights can potentially lead to possible constraints on human motions which may be used to more effectively solve open problems in the computer graphics community, most notably the problem of (human) motion adaptation.
With the speedy increase of display resolution and the demand for interactive frame rate, rendering acceleration is becoming more critical for a wide range of virtual reality applications. Foveated rendering addresses...
详细信息
ISBN:
(纸本)9781665496179
With the speedy increase of display resolution and the demand for interactive frame rate, rendering acceleration is becoming more critical for a wide range of virtual reality applications. Foveated rendering addresses this challenge by rendering with a non-uniform resolution for the display. Motivated by the non-linear optical lens equation, we present rectangular mapping-based foveated rendering (RMFR), a simple yet effective implementation of foveated rendering framework. RMFR supports varying level of foveation according to the eccentricity and the scene complexity. Compared with traditional foveated rendering methods, rectangular mapping-based foveated rendering provides a superior level of perceived visual quality while consuming minimal rendering cost.
Social priming is the idea that observations of a virtual human (VH) engaged in short social interactions with a real or virtual human bystander can positively influence users' subsequent interactions with that VH...
详细信息
ISBN:
(纸本)9781665496179
Social priming is the idea that observations of a virtual human (VH) engaged in short social interactions with a real or virtual human bystander can positively influence users' subsequent interactions with that VH. In this paper we investigate the question of whether the positive effects of social priming are limited to interactions with humanoid entities. For instance, virtual dogs offer an attractive candidate for non-humanoid entities, as previous research suggests multiple positive effects. In particular, real human dog owners receive more positive attention from strangers than non-dog owners. To examine the influence of such social priming we carried out a human-subjects experiment with four conditions: three social priming conditions where a participant initially observed a VH interacting with one of three virtual entities (another VH, a virtual pet dog, or a virtual personal robot), and a non-social priming condition where a VH (alone) was intently looking at her phone as if reading something. We recruited 24 participants and conducted a mixed-methods analysis. We found that a VH's prior social interactions with another VH and a virtual dog significantly increased participants' perceptions of the VHs' affective attraction. Also, participants felt more inclined to interact with the VH in the future in all of the social priming conditions. Qualitatively, we found that the social priming conditions resulted in a more positive user experience than the non-social priming condition. Also, the virtual dog and the virtual robot were perceived as a source of positive surprise, with participants appreciating the non-humanoid interactions for various reasons, such as the avoidance of social anxieties sometimes associated with humans.
In recent years, research has been conducted on the reflection of real space onto virtual objects. However, conventional methods have not been able to achieve accurate reflection due to problems such as the direction ...
详细信息
ISBN:
(数字)9781510653320
ISBN:
(纸本)9781510653320;9781510653313
In recent years, research has been conducted on the reflection of real space onto virtual objects. However, conventional methods have not been able to achieve accurate reflection due to problems such as the direction and angle of the reflected object not matching. In addition, since the light source is not estimated, the color of the virtual object floats away from the surroundings and shadows are not generated. In this study, we propose a method to integrate virtual objects and real space without any discomfort by using light source estimation.
In this paper, a semi-analytical surface tension model is proposed for smoothed particle hydrodynamics (SPH). Different from previous approaches, cohesive and adhesive forces in our model are unified within a surface ...
详细信息
ISBN:
(纸本)9781665484022
In this paper, a semi-analytical surface tension model is proposed for smoothed particle hydrodynamics (SPH). Different from previous approaches, cohesive and adhesive forces in our model are unified within a surface energy framework for nonuniform systems. To calculate the adhesive force, we use a semi-analytical solution to convert the volume integral into a surface integral, triangular meshes which represent solid boundaries can be directly introduced into liquid-solid interactions. A gradient descent algorithm is employed to optimize the objective function, which represents the total energy of the fluid. Experiments show that our model can efficiently handle complex solid boundaries with surface-tension-driven phenomena.
In this study, we propose an approach to display a distortion-free mid-air image inside a transparent refractive object and on a curved reflective surface. We compensate for the distortion by generating a light source...
详细信息
ISBN:
(纸本)9781665496179
In this study, we propose an approach to display a distortion-free mid-air image inside a transparent refractive object and on a curved reflective surface. We compensate for the distortion by generating a light source image that cancels the distortions in the mid-air image caused by refraction and reflection. The light source image is generated via ray-tracing simulation by transmitting the desired view image to the mid-air imaging system, which includes distortive surfaces, and by receiving the transmitted image at the light source position. The proposed methods can be applied to dynamic images using a light source image as a UV map in texture mapping. Finally, we present the results of an evaluation of our method performed in an actual optical system using the generated light source image, which visually demonstrate the effectiveness of the proposed approach.
Since for most consumers the Virtual Reality (VR) experience exceeds that of desktop applications, an increasing number of applications is being transferred from desktop to VR. Industrial and entertainment application...
详细信息
ISBN:
(纸本)9781665496179
Since for most consumers the Virtual Reality (VR) experience exceeds that of desktop applications, an increasing number of applications is being transferred from desktop to VR. Industrial and entertainment applications primarily expect for a richer consumer experience, while others, such as surgical applications, seek for improved precision over their desktop counterparts. One way to improve the performance of precision-based VR applications is to provide suitable visualizations. Today, these "suitable" visualizations are mostly transferred from desktop to VR without considering their spatial and temporal performance might change in VR. This may not lead to an optimal solution, which can be crucial for precision-based tasks. Misinterpretation of shape or distance in a surgical or pre-operative simulation can affect the chosen treatment and thus directly impact the outcome. Therefore, we evaluate the performance differences of multiple visualizations for 3D surfaces based on their shape and distance estimation for desktop and VR applications. We conducted a quantitative user study with 56 participants evaluating seven visualizations (Phong, Toon, Fresnel, Pseudo-Chromadepth, Heatmap, Isolines, and Arrow Glyphs). Our results show that the performance of each visualization varies depending on the task, system, and surface type, with VR generally providing improved results. While Isolines are able to improve distance estimation, Phong and Heatmaps are beneficial for shape estimation.
暂无评论