Shuang Wu

Chao Shuai, Jieming Zhong, Shuang Wu et al.

Deepfake has taken the world by storm, triggering a trust crisis. Current deepfake detection methods are typically inadequate in generalizability, with a tendency to overfit to image contents such as the background, which are frequently occurring but relatively unimportant in the training dataset. Furthermore, current methods heavily rely on a few dominant forgery regions and may ignore other equally important regions, leading to inadequate uncovering of forgery cues. In this paper, we strive to address these shortcomings from three aspects: (1) We propose an innovative two-stream network that effectively enlarges the potential regions from which the model extracts forgery evidence. (2) We devise three functional modules to handle the multi-stream and multi-scale features in a collaborative learning scheme. (3) Confronted with the challenge of obtaining forgery annotations, we propose a Semi-supervised Patch Similarity Learning strategy to estimate patch-level forged location annotations. Empirically, our method demonstrates significantly improved robustness and generalizability, outperforming previous methods on six benchmarks, and improving the frame-level AUC on Deepfake Detection Challenge preview dataset from 0.797 to 0.835 and video-level AUC on CelebDF$\_$v1 dataset from 0.811 to 0.847. Our implementation is available at https://github.com/sccsok/Locate-and-Verify.

Shuang Wu, Xiaoqiang Ren, Qing-Shan Jia et al.

We consider optimal sensor scheduling with unknown communication channel statistics. We formulate two types of scheduling problems with the communication rate being a soft or hard constraint, respectively. We first present some structural results on the optimal scheduling policy using dynamic programming and assuming the channel statistics is known. We prove that the Q-factor is monotonic and submodular, which leads to the threshold-like structures in both types of problems. Then we develop a stochastic approximation and parameter learning frameworks to deal with the two scheduling problems with unknown channel statistics. We utilize their structures to design specialized learning algorithms. We prove the convergence of these algorithms. Performance improvement compared with the standard Q-learning algorithm is shown through numerical examples.

Shuang Wu, Kemi Ding, Peng Cheng et al.

This work considers the sensor scheduling for multiple dynamic processes. We consider $n$ linear dynamic processes, the state of each process is measured by a sensor, which transmits their local state estimates over wireless channels to a remote estimator with certain communication costs. In each time step, only a portion of the sensors is allowed to transmit data to the remote estimator and the packet might be lost due to unreliability of the wireless channels. Our goal is to find a scheduling policy which coordinates the sensors in a centralized manner to minimize the total expected estimation error of the remote estimator and the communication costs. We formulate the problem as a Markov decision process. We develop an algorithm to check whether there exists a deterministic stationary optimal policy. We show the optimality of monotone policies, which saves computational effort of finding an optimal policy and facilitates practical implementation. Nevertheless, obtaining an exact optimal policy still suffers from curse of dimensionality when the number of processes are large. We further provide an index-based heuristics to avoid brute force computation. Numerical examples are presented to illustrate our theoretical results.

Shuang Wu, Xiaoqiang Ren, Subhrakanti Dey et al.

This paper considers the problem of sensory data scheduling of multiple processes. There are $n$ independent linear time-invariant processes and a remote estimator monitoring all the processes. Each process is measured by a sensor, which sends its local state estimate to the remote estimator. The sizes of the packets are different due to different dimensions of each process, and thus it may take different lengths of time steps for the sensors to send their data. Because of bandwidth limitation, only a portion of all the sensors are allowed to transmit. Our goal is to minimize the average of estimation error covariance of the whole system at the remote estimator. The problem is formulated as a Markov decision process (MDP) with average cost over an infinite time horizon. We prove the existence of a deterministic and stationary policy for the problem. We also find that the optimal policy has a consistent behavior and threshold type structure. A numerical example is provided to illustrate our main results.

Shuang Wu, Xiaoqiang Ren, Yiguang Hong et al.

We consider the design of a fair sensor schedule for a number of sensors monitoring different linear time-invariant processes. The largest average remote estimation error among all processes is to be minimized. We first consider a general setup for the max-min fair allocation problem. By reformulating the problem as its equivalent form, we transform the fair resource allocation problem into a zero-sum game between a "judge" and a resource allocator. We propose an equilibrium seeking procedure and show that there exists a unique Nash equilibrium in pure strategy for this game. We then apply the result to the sensor scheduling problem and show that the max-min fair sensor scheduling policy can be achieved.

Shuang Wu, Youtian Lin, Feihu Zhang et al.

Generating high-quality 3D assets from text and images has long been challenging, primarily due to the absence of scalable 3D representations capable of capturing intricate geometry distributions. In this work, we introduce Direct3D, a native 3D generative model scalable to in-the-wild input images, without requiring a multiview diffusion model or SDS optimization. Our approach comprises two primary components: a Direct 3D Variational Auto-Encoder (D3D-VAE) and a Direct 3D Diffusion Transformer (D3D-DiT). D3D-VAE efficiently encodes high-resolution 3D shapes into a compact and continuous latent triplane space. Notably, our method directly supervises the decoded geometry using a semi-continuous surface sampling strategy, diverging from previous methods relying on rendered images as supervision signals. D3D-DiT models the distribution of encoded 3D latents and is specifically designed to fuse positional information from the three feature maps of the triplane latent, enabling a native 3D generative model scalable to large-scale 3D datasets. Additionally, we introduce an innovative image-to-3D generation pipeline incorporating semantic and pixel-level image conditions, allowing the model to produce 3D shapes consistent with the provided conditional image input. Extensive experiments demonstrate the superiority of our large-scale pre-trained Direct3D over previous image-to-3D approaches, achieving significantly better generation quality and generalization ability, thus establishing a new state-of-the-art for 3D content creation. Project page: https://nju-3dv.github.io/projects/Direct3D/.

Shuang Wu, Xiaoning Song, Zhenhua Feng

Recently, word enhancement has become very popular for Chinese Named Entity Recognition (NER), reducing segmentation errors and increasing the semantic and boundary information of Chinese words. However, these methods tend to ignore the information of the Chinese character structure after integrating the lexical information. Chinese characters have evolved from pictographs since ancient times, and their structure often reflects more information about the characters. This paper presents a novel Multi-metadata Embedding based Cross-Transformer (MECT) to improve the performance of Chinese NER by fusing the structural information of Chinese characters. Specifically, we use multi-metadata embedding in a two-stream Transformer to integrate Chinese character features with the radical-level embedding. With the structural characteristics of Chinese characters, MECT can better capture the semantic information of Chinese characters for NER. The experimental results obtained on several well-known benchmarking datasets demonstrate the merits and superiority of the proposed MECT method.\footnote{The source code of the proposed method is publicly available at https://github.com/CoderMusou/MECT4CNER.

Shuang Wu, Youtian Lin, Feihu Zhang et al.

Generating high-resolution 3D shapes using volumetric representations such as Signed Distance Functions (SDFs) presents substantial computational and memory challenges. We introduce Direct3D-S2, a scalable 3D generation framework based on sparse volumes that achieves superior output quality with dramatically reduced training costs. Our key innovation is the Spatial Sparse Attention (SSA) mechanism, which greatly enhances the efficiency of Diffusion Transformer (DiT) computations on sparse volumetric data. SSA allows the model to effectively process large token sets within sparse volumes, substantially reducing computational overhead and achieving a 3.9x speedup in the forward pass and a 9.6x speedup in the backward pass. Our framework also includes a variational autoencoder (VAE) that maintains a consistent sparse volumetric format across input, latent, and output stages. Compared to previous methods with heterogeneous representations in 3D VAE, this unified design significantly improves training efficiency and stability. Our model is trained on public available datasets, and experiments demonstrate that Direct3D-S2 not only surpasses state-of-the-art methods in generation quality and efficiency, but also enables training at 1024 resolution using only 8 GPUs, a task typically requiring at least 32 GPUs for volumetric representations at 256 resolution, thus making gigascale 3D generation both practical and accessible. Project page: https://www.neural4d.com/research/direct3d-s2.