This paper targets high-fidelity and real-time view synthesis of dynamic 3D scenes at 4K resolution. Recently, some methods on dynamic view synthesis have shown impressive rendering quality. However, their speed is still limited when rendering high-resolution images. To overcome this problem, we propose 4K4D, a 4D point cloud representation that supports hardware rasterization and enables unprecedented rendering speed. Our representation is built on a 4D feature grid so that the points are naturally regularized and can be robustly optimized. In addition, we design a novel hybrid appearance model that significantly boosts the rendering quality while preserving efficiency. Moreover, we develop a differentiable depth peeling algorithm to effectively learn the proposed model from RGB videos. Experiments show that our representation can be rendered at over 400 FPS on the DNA-Rendering dataset at 1080p resolution and 80 FPS on the ENeRF-Outdoor dataset at 4K resolution using an RTX 4090 GPU, which is 30x faster than previous methods and achieves the state-of-the-art rendering quality. Our project page is available at https://zju3dv.github.io/4k4d/.
Video prediction aims to predict future frames from a video's previous content. Existing methods mainly process video data where the time dimension mingles with the space and channel dimensions from three distinct angles: as a sequence of individual frames, as a 3D volume in spatiotemporal coordinates, or as a stacked image where frames are treated as separate channels. Most of them generally focus on one of these perspectives and may fail to fully exploit the relationships across different dimensions. To address this issue, this paper introduces a convolutional mixer for video prediction, termed ViP-Mixer, to model the spatiotemporal evolution in the latent space of an autoencoder. The ViP-Mixers are stacked sequentially and interleave feature mixing at three levels: frames, channels, and locations. Extensive experiments demonstrate that our proposed method achieves new state-of-the-art prediction performance on three benchmark video datasets covering both synthetic and real-world scenarios.
To enable realistic experience in AR/VR and digital entertainment, we present the first point-based human avatar model that embodies the entirety expressive range of digital humans. We employ two MLPs to model pose-dependent deformation and linear skinning (LBS) weights. The representation of appearance relies on a decoder and the features that attached to each point. In contrast to alternative implicit approaches, the oriented points representation not only provides a more intuitive way to model human avatar animation but also significantly reduces both training and inference time. Moreover, we propose a novel method to transfer semantic information from the SMPL-X model to the points, which enables to better understand human body movements. By leveraging the semantic information of points, we can facilitate virtual try-on and human avatar composition through exchanging the points of same category across different subjects. Experimental results demonstrate the efficacy of our presented method.
In last decades, dynamic resource programming in partial resource domains has been extensively investigated for single time slot optimizations. However, with the emerging real-time media applications in fifth-generation communications, their new quality of service requirements are often measured in temporal dimension. This requires multistage optimization for full resource domain dynamic programming. Taking experience rate as a typical temporal multistage metric, we jointly optimize time, frequency, space and power domains resource for multistage optimization. To strike a good tradeoff between system performance and computational complexity, we first transform the formulated mixed integer non-linear constraints into equivalent convex second order cone constraints, by exploiting the coupling effect among the resources. Leveraging the concept of structural sparsity, the objective of max-min experience rate is given as a weighted 1-norm term associated with the precoding matrix. Finally, a low-complexity iterative algorithm is proposed for full resource domain programming, aided by another simple conic optimization for obtaining its feasible initial result. Simulation verifies that our design significantly outperform the benchmarks while maintaining a fast convergence rate, shedding light on full domain dynamic resource programming of multistage optimizations.
Reconstructing real-world 3D objects has numerous applications in computer vision, such as virtual reality, video games, and animations. Ideally, 3D reconstruction methods should generate high-fidelity results with 3D consistency in real-time. Traditional methods match pixels between images using photo-consistency constraints or learned features, while differentiable rendering methods like Neural Radiance Fields (NeRF) use differentiable volume rendering or surface-based representation to generate high-fidelity scenes. However, these methods require excessive runtime for rendering, making them impractical for daily applications. To address these challenges, we present $\textbf{EvaSurf}$, an $\textbf{E}$fficient $\textbf{V}$iew-$\textbf{A}$ware implicit textured $\textbf{Surf}$ace reconstruction method on mobile devices. In our method, we first employ an efficient surface-based model with a multi-view supervision module to ensure accurate mesh reconstruction. To enable high-fidelity rendering, we learn an implicit texture embedded with a set of Gaussian lobes to capture view-dependent information. Furthermore, with the explicit geometry and the implicit texture, we can employ a lightweight neural shader to reduce the expense of computation and further support real-time rendering on common mobile devices. Extensive experiments demonstrate that our method can reconstruct high-quality appearance and accurate mesh on both synthetic and real-world datasets. Moreover, our method can be trained in just 1-2 hours using a single GPU and run on mobile devices at over 40 FPS (Frames Per Second), with a final package required for rendering taking up only 40-50 MB.
Bayesian neural networks use random variables to describe the neural networks rather than deterministic neural networks and are mostly trained by variational inference which updates the mean and variance at the same time. Here, we formulate the Bayesian neural networks as a minimax game problem. We do the experiments on the MNIST data set and the primary result is comparable to the existing closed-loop transcription neural network. Finally, we reveal the connections between Bayesian neural networks and closed-loop transcription neural networks, and show our framework is rather practical, and provide another view of Bayesian neural networks.
We study the sample complexity of learning an $\varepsilon$-optimal policy in an average-reward Markov decision process (MDP) under a generative model. We establish the complexity bound $\widetilde{O}\left(SA\frac{H}{\varepsilon^2} \right)$, where $H$ is the span of the bias function of the optimal policy and $SA$ is the cardinality of the state-action space. Our result is the first that is minimax optimal (up to log factors) in all parameters $S,A,H$ and $\varepsilon$, improving on existing work that either assumes uniformly bounded mixing times for all policies or has suboptimal dependence on the parameters. Our result is based on reducing the average-reward MDP to a discounted MDP. To establish the optimality of this reduction, we develop improved bounds for $\gamma$-discounted MDPs, showing that $\widetilde{O}\left(SA\frac{H}{(1-\gamma)^2\varepsilon^2} \right)$ samples suffice to learn a $\varepsilon$-optimal policy in weakly communicating MDPs under the regime that $\gamma \geq 1 - \frac{1}{H}$, circumventing the well-known lower bound of $\widetilde{\Omega}\left(SA\frac{1}{(1-\gamma)^3\varepsilon^2} \right)$ for general $\gamma$-discounted MDPs. Our analysis develops upper bounds on certain instance-dependent variance parameters in terms of the span parameter. These bounds are tighter than those based on the mixing time or diameter of the MDP and may be of broader use.
Deep models deployed on edge devices frequently encounter resource variability, which arises from fluctuating energy levels, timing constraints, or prioritization of other critical tasks within the system. State-of-the-art machine learning pipelines generate resource-agnostic models, not capable to adapt at runtime. In this work we introduce Resource-Efficient Deep Subnetworks (REDS) to tackle model adaptation to variable resources. In contrast to the state-of-the-art, REDS use structured sparsity constructively by exploiting permutation invariance of neurons, which allows for hardware-specific optimizations. Specifically, REDS achieve computational efficiency by (1) skipping sequential computational blocks identified by a novel iterative knapsack optimizer, and (2) leveraging simple math to re-arrange the order of operations in REDS computational graph to take advantage of the data cache. REDS support conventional deep networks frequently deployed on the edge and provide computational benefits even for small and simple networks. We evaluate REDS on six benchmark architectures trained on the Google Speech Commands, FMNIST and CIFAR10 datasets, and test on four off-the-shelf mobile and embedded hardware platforms. We provide a theoretical result and empirical evidence for REDS outstanding performance in terms of submodels' test set accuracy, and demonstrate an adaptation time in response to dynamic resource constraints of under 40$\mu$s, utilizing a 2-layer fully-connected network on Arduino Nano 33 BLE Sense.
A multi-modal emotion recognition method was established by combining two-channel convolutional neural network with ring network. This method can extract emotional information effectively and improve learning efficiency. The words were vectorized with GloVe, and the word vector was input into the convolutional neural network. Combining attention mechanism and maximum pool converter BiSRU channel, the local deep emotion and pre-post sequential emotion semantics are obtained. Finally, multiple features are fused and input as the polarity of emotion, so as to achieve the emotion analysis of the target. Experiments show that the emotion analysis method based on feature fusion can effectively improve the recognition accuracy of emotion data set and reduce the learning time. The model has a certain generalization.
Within the ever-evolving landscape of engineering, particularly in the dynamic domain of additive In manufacturing, a pursuit of precision and excellence in production processes takes centre stage. This research , This paper serves to give a comprehensive understanding of piezoelectric sensors, a topic that is both academically engaging and of practical significance, catering to both seasoned experts and those newly venturing into the field. Additive manufacturing, lauded for its groundbreaking potential, underscores the imperative of rigorous quality control. This introduces piezoelectric sensors, devices that may be unfamiliar to many but possess considerable potential. This paper embarks on a methodical journey, commencing with an introductory elucidation of the piezoelectric effect. It then advances to the vital role of piezoelectric sensors in real-time monitoring and quality control, unveiling their potential and relevance for newcomers and seasoned professionals alike. This research, structured systematically from fundamental principles to pragmatic applications, presents findings that are not only academically informative but also represent a substantial stride towards achieving precision and high-quality manufacturing processes in the engineering field.