Xiang Cao

Xu Gu, Yuchong Sun, Feiyue Ni et al.

A video storyboard is a roadmap for video creation which consists of shot-by-shot images to visualize key plots in a text synopsis. Creating video storyboards, however, remains challenging which not only requires cross-modal association between high-level texts and images but also demands long-term reasoning to make transitions smooth across shots. In this paper, we propose a new task called Text synopsis to Video Storyboard (TeViS) which aims to retrieve an ordered sequence of images as the video storyboard to visualize the text synopsis. We construct a MovieNet-TeViS dataset based on the public MovieNet dataset. It contains 10K text synopses each paired with keyframes manually selected from corresponding movies by considering both relevance and cinematic coherence. To benchmark the task, we present strong CLIP-based baselines and a novel VQ-Trans. VQ-Trans first encodes text synopsis and images into a joint embedding space and uses vector quantization (VQ) to improve the visual representation. Then, it auto-regressively generates a sequence of visual features for retrieval and ordering. Experimental results demonstrate that VQ-Trans significantly outperforms prior methods and the CLIP-based baselines. Nevertheless, there is still a large gap compared to human performance suggesting room for promising future work. The code and data are available at: \url{https://ruc-aimind.github.io/projects/TeViS/}

Liangqi Zhang, Haibo Shen, Yihao Luo et al.

Modern efficient Convolutional Neural Networks(CNNs) always use Depthwise Separable Convolutions(DSCs) and Neural Architecture Search(NAS) to reduce the number of parameters and the computational complexity. But some inherent characteristics of networks are overlooked. Inspired by visualizing feature maps and N$\times$N(N$>$1) convolution kernels, several guidelines are introduced in this paper to further improve parameter efficiency and inference speed. Based on these guidelines, our parameter-efficient CNN architecture, called \textit{VGNetG}, achieves better accuracy and lower latency than previous networks with about 30%$\thicksim$50% parameters reduction. Our VGNetG-1.0MP achieves 67.7% top-1 accuracy with 0.99M parameters and 69.2% top-1 accuracy with 1.14M parameters on ImageNet classification dataset. Furthermore, we demonstrate that edge detectors can replace learnable depthwise convolution layers to mix features by replacing the N$\times$N kernels with fixed edge detection kernels. And our VGNetF-1.5MP archives 64.4%(-3.2%) top-1 accuracy and 66.2%(-1.4%) top-1 accuracy with additional Gaussian kernels.

Haibo Shen, Juyu Xiao, Yihao Luo et al.

Neuromorphic vision sensors (event cameras) simulate biological visual perception systems and have the advantages of high temporal resolution, less data redundancy, low power consumption, and large dynamic range. Since both events and spikes are modeled from neural signals, event cameras are inherently suitable for spiking neural networks (SNNs), which are considered promising models for artificial intelligence (AI) and theoretical neuroscience. However, the unconventional visual signals of these cameras pose a great challenge to the robustness of spiking neural networks. In this paper, we propose a novel data augmentation method, ViewPoint Transform and SpatioTemporal Stretching (VPT-STS). It improves the robustness of SNNs by transforming the rotation centers and angles in the spatiotemporal domain to generate samples from different viewpoints. Furthermore, we introduce the spatiotemporal stretching to avoid potential information loss in viewpoint transformation. Extensive experiments on prevailing neuromorphic datasets demonstrate that VPT-STS is broadly effective on multi-event representations and significantly outperforms pure spatial geometric transformations. Notably, the SNNs model with VPT-STS achieves a state-of-the-art accuracy of 84.4\% on the DVS-CIFAR10 dataset.

Haibo Shen, Yihao Luo, Xiang Cao et al.

Neuromorphic vision sensors (event cameras) are inherently suitable for spiking neural networks (SNNs) and provide novel neuromorphic vision data for this biomimetic model. Due to the spatiotemporal characteristics, novel data augmentations are required to process the unconventional visual signals of these cameras. In this paper, we propose a novel Event SpatioTemporal Fragments (ESTF) augmentation method. It preserves the continuity of neuromorphic data by drifting or inverting fragments of the spatiotemporal event stream to simulate the disturbance of brightness variations, leading to more robust spiking neural networks. Extensive experiments are performed on prevailing neuromorphic datasets. It turns out that ESTF provides substantial improvements over pure geometric transformations and outperforms other event data augmentation methods. It is worth noting that the SNNs with ESTF achieve the state-of-the-art accuracy of 83.9\% on the CIFAR10-DVS dataset.

Haibo Shen, Yihao Luo, Xiang Cao et al.

As the third generation of neural networks, spiking neural networks (SNNs) are dedicated to exploring more insightful neural mechanisms to achieve near-biological intelligence. Intuitively, biomimetic mechanisms are crucial to understanding and improving SNNs. For example, the associative long-term potentiation (ALTP) phenomenon suggests that in addition to learning mechanisms between neurons, there are associative effects within neurons. However, most existing methods only focus on the former and lack exploration of the internal association effects. In this paper, we propose a novel Adaptive Internal Association~(AIA) neuron model to establish previously ignored influences within neurons. Consistent with the ALTP phenomenon, the AIA neuron model is adaptive to input stimuli, and internal associative learning occurs only when both dendrites are stimulated at the same time. In addition, we employ weighted weights to measure internal associations and introduce intermediate caches to reduce the volatility of associations. Extensive experiments on prevailing neuromorphic datasets show that the proposed method can potentiate or depress the firing of spikes more specifically, resulting in better performance with fewer spikes. It is worth noting that without adding any parameters at inference, the AIA model achieves state-of-the-art performance on DVS-CIFAR10~(83.9\%) and N-CARS~(95.64\%) datasets.

Liangqi Zhang, Yihao Luo, Xiang Cao et al.

Convolutional neural networks (CNNs) have achieved superior performance but still lack clarity about the nature and properties of feature extraction. In this paper, by analyzing the sensitivity of neural networks to frequencies and scales, we find that neural networks not only have low- and medium-frequency biases but also prefer different frequency bands for different classes, and the scale of objects influences the preferred frequency bands. These observations lead to the hypothesis that neural networks must learn the ability to extract features at various scales and frequencies. To corroborate this hypothesis, we propose a network architecture based on Gaussian derivatives, which extracts features by constructing scale space and employing partial derivatives as local feature extraction operators to separate high-frequency information. This manually designed method of extracting features from different scales allows our GSSDNets to achieve comparable accuracy with vanilla networks on various datasets.

Tiwei Bie, Maosong Cao, Xiang Cao et al.

While LLaDA2.0 showcased the scaling potential of 100B-level block-diffusion models and their inherent parallelization, the delicate equilibrium between decoding speed and generation quality has remained an elusive frontier. Today, we unveil LLaDA2.1, a paradigm shift designed to transcend this trade-off. By seamlessly weaving Token-to-Token (T2T) editing into the conventional Mask-to-Token (M2T) scheme, we introduce a joint, configurable threshold-decoding scheme. This structural innovation gives rise to two distinct personas: the Speedy Mode (S Mode), which audaciously lowers the M2T threshold to bypass traditional constraints while relying on T2T to refine the output; and the Quality Mode (Q Mode), which leans into conservative thresholds to secure superior benchmark performances with manageable efficiency degrade. Furthering this evolution, underpinned by an expansive context window, we implement the first large-scale Reinforcement Learning (RL) framework specifically tailored for dLLMs, anchored by specialized techniques for stable gradient estimation. This alignment not only sharpens reasoning precision but also elevates instruction-following fidelity, bridging the chasm between diffusion dynamics and complex human intent. We culminate this work by releasing LLaDA2.1-Mini (16B) and LLaDA2.1-Flash (100B). Across 33 rigorous benchmarks, LLaDA2.1 delivers strong task performance and lightning-fast decoding speed. Despite its 100B volume, on coding tasks it attains an astounding 892 TPS on HumanEval+, 801 TPS on BigCodeBench, and 663 TPS on LiveCodeBench.

Xiang Cao, Qiaoqiao Ding, Xinliang Liu et al.

Ultrasound Computed Tomography (USCT) constitutes a nonlinear inverse problem with inherent ill-posedness that can benefit from regularization through diffusion generative priors. However, traditional approaches for solving Helmholtz equation-constrained USCT face three fundamental challenges when integrating these priors: PDE-constrained gradient computation, discretization-induced approximation errors, and computational imbalance between neural networks and numerical PDE solvers. In this work, we introduce \textbf{Diff-ANO} (\textbf{Diff}usion-based Models with \textbf{A}djoint \textbf{N}eural \textbf{O}perators), a novel framework that combines conditional consistency models with adjoint operator learning to address these limitations. Our two key innovations include: (1) a \textit{conditional consistency model} that enables measurement-conditional few-step sampling by directly learning a self-consistent mapping from diffusion trajectories, and (2) an \textit{adjoint operator learning} module that replaces traditional PDE solvers with neural operator surrogates for efficient adjoint-based gradient computation. To enable practical deployment, we introduce the batch-based Convergent Born Series (BCBS)--a memory-efficient strategy for online generation of neural operator training pairs. Comprehensive experiments demonstrate that Diff-ANO significantly improves both computational efficiency and reconstruction quality, especially under sparse-view and partial-view measurement scenarios.

Xurui Li, Xiang Cao, Xuetao Qiu et al.

Criminals are using every means available to launder the profits from their illegal activities into ostensibly legitimate assets. Meanwhile, most commercial anti-money laundering systems are still rule-based, which cannot adapt to the ever-changing tricks. Although some machine learning methods have been proposed, they are mainly focused on the perspective of abnormal behavior for single accounts. Considering money laundering activities are often involved in gang criminals, these methods are still not intelligent enough to crack down on criminal gangs all-sidedly. In this paper, a systematic solution is presented to find suspicious money laundering gangs. A temporal-directed Louvain algorithm has been proposed to detect communities according to relevant anti-money laundering patterns. All processes are implemented and optimized on Spark platform. This solution can greatly improve the efficiency of anti-money laundering work for financial regulation agencies.

Xiang Cao, Qiaoqiao Ding, Xiaoqun Zhang

The regularized D-bar method is a popular method for solving Electrical Impedance Tomography (EIT) problems due to its efficiency and simplicity. It utilizes the low-pass truncated scattering data in the non-linear Fourier domain to solve the associated D-bar integral equations, yielding a smooth conductivity approximation. However, the D-bar reconstruction often presents low contrast and resolution due to the absence of accurate high-frequency information and the ill-posedness of the problem. In this paper, we propose a deep learning-based supervised approach for real-time EIT reconstruction. Based on the D-bar method, we propose to utilize both multi-scale frequency enhancement and spatial consistency for a high image quality reconstruction. Additionally, we propose a fixed-point iteration for solving discrete D-bar systems on GPUs for fast computation. Numerical results are performed for both the continuum model and complete electrode model simulation on KIT4 and ACT4 datasets to demonstrate notable improvements in absolute EIT imaging quality.

Yihao Luo, Xiang Cao, Juntao Zhang et al.

With the continuous improvement of the performance of object detectors via advanced model architectures, imbalance problems in the training process have received more attention. It is a common paradigm in object detection frameworks to perform multi-scale detection. However, each scale is treated equally during training. In this paper, we carefully study the objective imbalance of multi-scale detector training. We argue that the loss in each scale level is neither equally important nor independent. Different from the existing solutions of setting multi-task weights, we dynamically optimize the loss weight of each scale level in the training process. Specifically, we propose an Adaptive Variance Weighting (AVW) to balance multi-scale loss according to the statistical variance. Then we develop a novel Reinforcement Learning Optimization (RLO) to decide the weighting scheme probabilistically during training. The proposed dynamic methods make better utilization of multi-scale training loss without extra computational complexity and learnable parameters for backpropagation. Experiments show that our approaches can consistently boost the performance over various baseline detectors on Pascal VOC and MS COCO benchmark.

Yihao Luo, Xiang Cao, Juntao Zhang et al.

Feature pyramid network (FPN) has been an effective framework to extract multi-scale features in object detection. However, current FPN-based methods mostly suffer from the intrinsic flaw of channel reduction, which brings about the loss of semantical information. And the miscellaneous fused feature maps may cause serious aliasing effects. In this paper, we present a novel channel enhancement feature pyramid network (CE-FPN) with three simple yet effective modules to alleviate these problems. Specifically, inspired by sub-pixel convolution, we propose a sub-pixel skip fusion method to perform both channel enhancement and upsampling. Instead of the original 1x1 convolution and linear upsampling, it mitigates the information loss due to channel reduction. Then we propose a sub-pixel context enhancement module for extracting more feature representations, which is superior to other context methods due to the utilization of rich channel information by sub-pixel convolution. Furthermore, a channel attention guided module is introduced to optimize the final integrated features on each level, which alleviates the aliasing effect only with a few computational burdens. Our experiments show that CE-FPN achieves competitive performance compared to state-of-the-art FPN-based detectors on MS COCO benchmark.

Yihao Luo, Min Xu, Caihong Yuan et al.

Recently spiking neural networks (SNNs), the third-generation of neural networks has shown remarkable capabilities of energy-efficient computing, which is a promising alternative for deep neural networks (DNNs) with high energy consumption. SNNs have reached competitive results compared to DNNs in relatively simple tasks and small datasets such as image classification and MNIST/CIFAR, while few studies on more challenging vision tasks on complex datasets. In this paper, we focus on extending deep SNNs to object tracking, a more advanced vision task with embedded applications and energy-saving requirements, and present a spike-based Siamese network called SiamSNN. Specifically, we propose an optimized hybrid similarity estimation method to exploit temporal information in the SNNs, and introduce a novel two-status coding scheme to optimize the temporal distribution of output spike trains for further improvements. SiamSNN is the first deep SNN tracker that achieves short latency and low precision loss on the visual object tracking benchmarks OTB2013/2015, VOT2016/2018, and GOT-10k. Moreover, SiamSNN achieves notably low energy consumption and real-time on Neuromorphic chip TrueNorth.

Yi Xiao, Xiang Cao, Xianyi Zhu et al.

High-resolution depth map can be inferred from a low-resolution one with the guidance of an additional high-resolution texture map of the same scene. Recently, deep neural networks with large receptive fields are shown to benefit applications such as image completion. Our insight is that super resolution is similar to image completion, where only parts of the depth values are precisely known. In this paper, we present a joint convolutional neural pyramid model with large receptive fields for joint depth map super-resolution. Our model consists of three sub-networks, two convolutional neural pyramids concatenated by a normal convolutional neural network. The convolutional neural pyramids extract information from large receptive fields of the depth map and guidance map, while the convolutional neural network effectively transfers useful structures of the guidance image to the depth image. Experimental results show that our model outperforms existing state-of-the-art algorithms not only on data pairs of RGB/depth images, but also on other data pairs like color/saliency and color-scribbles/colorized images.