Computational Visual Media最新文献

We introduce a novel end-to-end deep-learning solution for rapidly estimating a dense spherical depth map of an indoor environment. Our input is a single equirectangular image registered with a sparse depth map, as provided by a variety of common capture setups. Depth is inferred by an efficient and lightweight single-branch network, which employs a dynamic gating system to process together dense visual data and sparse geometric data. We exploit the characteristics of typical man-made environments to efficiently compress multi-resolution features and find short- and long-range relations among scene parts. Furthermore, we introduce a new augmentation strategy to make the model robust to different types of sparsity, including those generated by various structured light sensors and LiDAR setups. The experimental results demonstrate that our method provides interactive performance and outperforms state-of-the-art solutions in computational efficiency, adaptivity to variable depth sparsity patterns, and prediction accuracy for challenging indoor data, even when trained solely on synthetic data without any fine tuning.

We propose a unified 3D flow framework for joint learning of shape embedding and deformation for different categories. Our goal is to recover shapes from imperfect point clouds by fitting the best shape template in a shape repository after deformation. Accordingly, we learn a shape embedding for template retrieval and a flow-based network for robust deformation. We note that the deformation flow can be quite different for different shape categories. Therefore, we introduce a novel multi-hub module to learn multiple modes of deformation to incorporate such variation, providing a network which can handle a wide range of objects from different categories. The shape embedding is designed to retrieve the best-fit template as the nearest neighbor in a latent space. We replace the standard fully connected layer with a tiny structure in the embedding that significantly reduces network complexity and further improves deformation quality. Experiments show the superiority of our method to existing state-of-the-art methods via qualitative and quantitative comparisons. Finally, our method provides efficient and flexible deformation that can further be used for novel shape design.

Learning and inferring underlying motion patterns of captured 2D scenes and then re-creating dynamic evolution consistent with the real-world natural phenomena have high appeal for graphics and animation. To bridge the technical gap between virtual and real environments, we focus on the inverse modeling and reconstruction of visually consistent and property-verifiable oceans, taking advantage of deep learning and differentiable physics to learn geometry and constitute waves in a self-supervised manner. First, we infer hierarchical geometry using two networks, which are optimized via the differentiable renderer. We extract wave components from the sequence of inferred geometry through a network equipped with a differentiable ocean model. Then, ocean dynamics can be evolved using the reconstructed wave components. Through extensive experiments, we verify that our new method yields satisfactory results for both geometry reconstruction and wave estimation. Moreover, the new framework has the inverse modeling potential to facilitate a host of graphics applications, such as the rapid production of physically accurate scene animation and editing guided by real ocean scenes.

Visual simultaneous localisation and mapping (vSLAM) finds applications for indoor and outdoor navigation that routinely subjects it to visual complexities, particularly mirror reflections. The effect of mirror presence (time visible and its average size in the frame) was hypothesised to impact localisation and mapping performance, with systems using direct techniques expected to perform worse. Thus, a dataset, MirrEnv, of image sequences recorded in mirror environments, was collected, and used to evaluate the performance of existing representative methods. RGBD ORB-SLAM3 and BundleFusion appear to show moderate degradation of absolute trajectory error with increasing mirror duration, whilst the remaining results did not show significantly degraded localisation performance. The mesh maps generated proved to be very inaccurate, with real and virtual reflections colliding in the reconstructions. A discussion is given of the likely sources of error and robustness in mirror environments, outlining future directions for validating and improving vSLAM performance in the presence of planar mirrors. The MirrEnv dataset is available at https://doi.org/10.17035/d.2023.0292477898.

Free-viewpoint video allows the user to view objects from any virtual perspective, creating an immersive visual experience. This technology enhances the interactivity and freedom of multimedia performances. However, many free-viewpoint video synthesis methods hardly satisfy the requirement to work in real time with high precision, particularly for sports fields having large areas and numerous moving objects. To address these issues, we propose a free-viewpoint video synthesis method based on distance field acceleration. The central idea is to fuse multi-view distance field information and use it to adjust the search step size adaptively. Adaptive step size search is used in two ways: for fast estimation of multi-object three-dimensional surfaces, and synthetic view rendering based on global occlusion judgement. We have implemented our ideas using parallel computing for interactive display, using CUDA and OpenGL frameworks, and have used real-world and simulated experimental datasets for evaluation. The results show that the proposed method can render free-viewpoint videos with multiple objects on large sports fields at 25 fps. Furthermore, the visual quality of our synthetic novel viewpoint images exceeds that of state-of-the-art neural-rendering-based methods.

Visual object tracking has been drawing increasing attention in recent years, as a fundamental task in computer vision. To extend the range of tracking applications, researchers have been introducing information from multiple modalities to handle specific scenes, with promising research prospects for emerging methods and benchmarks. To provide a thorough review of multi-modal tracking, different aspects of multi-modal tracking algorithms are summarized under a unified taxonomy, with specific focus on visible-depth (RGB-D) and visible-thermal (RGB-T) tracking. Subsequently, a detailed description of the related benchmarks and challenges is provided. Extensive experiments were conducted to analyze the effectiveness of trackers on five datasets: PTB, VOT19-RGBD, GTOT, RGBT234, and VOT19-RGBT. Finally, various future directions, including model design and dataset construction, are discussed from different perspectives for further research.

We present a novel framework for the multi-domain synthesis of artworks from semantic layouts. One of the main limitations of this challenging task is the lack of publicly available segmentation datasets for art synthesis. To address this problem, we propose a dataset called ArtSem that contains 40,000 images of artwork from four different domains, with their corresponding semantic label maps. We first extracted semantic maps from landscape photography and used a conditional generative adversarial network (GAN)-based approach for generating high-quality artwork from semantic maps without requiring paired training data. Furthermore, we propose an artwork-synthesis model using domain-dependent variational encoders for high-quality multi-domain synthesis. Subsequently, the model was improved and complemented with a simple but effective normalization method based on jointly normalizing semantics and style, which we call spatially style-adaptive normalization (SSTAN). Compared to the previous methods, which only take semantic layout as the input, our model jointly learns style and semantic information representation, improving the generation quality of artistic images. These results indicate that our model learned to separate the domains in the latent space. Thus, we can perform fine-grained control of the synthesized artwork by identifying hyperplanes that separate the different domains. Moreover, by combining the proposed dataset and approach, we generated user-controllable artworks of higher quality than that of existing approaches, as corroborated by quantitative metrics and a user study.

Video colorization is a challenging and highly ill-posed problem. Although recent years have witnessed remarkable progress in single image colorization, there is relatively less research effort on video colorization, and existing methods always suffer from severe flickering artifacts (temporal inconsistency) or unsatisfactory colorization. We address this problem from a new perspective, by jointly considering colorization and temporal consistency in a unified framework. Specifically, we propose a novel temporally consistent video colorization (TCVC) framework. TCVC effectively propagates frame-level deep features in a bidirectional way to enhance the temporal consistency of colorization. Furthermore, TCVC introduces a self-regularization learning (SRL) scheme to minimize the differences in predictions obtained using different time steps. SRL does not require any ground-truth color videos for training and can further improve temporal consistency. Experiments demonstrate that our method can not only provide visually pleasing colorized video, but also with clearly better temporal consistency than state-of-the-art methods. A video demo is provided at https://www.youtube.com/watch?v=c7dczMs-olE, while code is available at https://github.com/lyh-18/TCVC-Temporally-Consistent-Video-Colorization.

Hierarchical multi-granularity image classification is a challenging task that aims to tag each given image with multiple granularity labels simultaneously. Existing methods tend to overlook that different image regions contribute differently to label prediction at different granularities, and also insufficiently consider relationships between the hierarchical multi-granularity labels. We introduce a sequence-to-sequence mechanism to overcome these two problems and propose a multi-granularity sequence generation (MGSG) approach for the hierarchical multi-granularity image classification task. Specifically, we introduce a transformer architecture to encode the image into visual representation sequences. Next, we traverse the taxonomic tree and organize the multi-granularity labels into sequences, and vectorize them and add positional information. The proposed multi-granularity sequence generation method builds a decoder that takes visual representation sequences and semantic label embedding as inputs, and outputs the predicted multi-granularity label sequence. The decoder models dependencies and correlations between multi-granularity labels through a masked multi-head self-attention mechanism, and relates visual information to the semantic label information through a cross-modality attention mechanism. In this way, the proposed method preserves the relationships between labels at different granularity levels and takes into account the influence of different image regions on labels with different granularities. Evaluations on six public benchmarks qualitatively and quantitatively demonstrate the advantages of the proposed method. Our project is available at https://github.com/liuxindazz/mgsg.

Data-driven garment animation is a current topic of interest in the computer graphics industry. Existing approaches generally establish the mapping between a single human pose or a temporal pose sequence, and garment deformation, but it is difficult to quickly generate diverse clothed human animations. We address this problem with a method to automatically synthesize dressed human animations with temporal consistency from a specified human motion label. At the heart of our method is a two-stage strategy. Specifically, we first learn a latent space encoding the sequence-level distribution of human motions utilizing a transformer-based conditional variational autoencoder (Transformer-CVAE). Then a garment simulator synthesizes dynamic garment shapes using a transformer encoder–decoder architecture. Since the learned latent space comes from varied human motions, our method can generate a variety of styles of motions given a specific motion label. By means of a novel beginning of sequence (BOS) learning strategy and a self-supervised refinement procedure, our garment simulator is capable of efficiently synthesizing garment deformation sequences corresponding to the generated human motions while maintaining temporal and spatial consistency. We verify our ideas experimentally. This is the first generative model that directly dresses human animation.