Dong Xu

Yiming Wu, Zhenghao Chen, Huan Wang et al.

The high computational cost and slow inference time are major obstacles to deploying Video Diffusion Models (VDMs). To overcome this, we introduce a new Video Diffusion Model Compression approach using individual content and motion dynamics preserved pruning and consistency loss. First, we empirically observe that deeper VDM layers are crucial for maintaining the quality of \textbf{motion dynamics} (\textit{e.g.,} coherence of the entire video), while shallower layers are more focused on \textbf{individual content} (\textit{e.g.,} individual frames). Therefore, we prune redundant blocks from the shallower layers while preserving more of the deeper layers, resulting in a lightweight VDM variant called VDMini. Moreover, we propose an \textbf{Individual Content and Motion Dynamics (ICMD)} Consistency Loss to gain comparable generation performance as larger VDM to VDMini. In particular, we first use the Individual Content Distillation (ICD) Loss to preserve the consistency in the features of each generated frame between the teacher and student models. Next, we introduce a Multi-frame Content Adversarial (MCA) Loss to enhance the motion dynamics across the generated video as a whole. This method significantly accelerates inference time while maintaining high-quality video generation. Extensive experiments demonstrate the effectiveness of our VDMini on two important video generation tasks, Text-to-Video (T2V) and Image-to-Video (I2V), where we respectively achieve an average 2.5 $\times$, 1.4 $\times$, and 1.25 $\times$ speed up for the I2V method SF-V, the T2V method T2V-Turbo-v2, and the T2V method HunyuanVideo, while maintaining the quality of the generated videos on several benchmarks including UCF101, VBench-T2V, and VBench-I2V.

Qin Zhou, Zhiyang Zhang, Jinglong Wang et al.

Diffusion models excel at image generation. Recent studies have shown that these models not only generate high-quality images but also encode text-image alignment information through attention maps or loss functions. This information is valuable for various downstream tasks, including segmentation, text-guided image editing, and compositional image generation. However, current methods heavily rely on the assumption of perfect text-image alignment in diffusion models, which is not the case. In this paper, we propose using zero-shot referring image segmentation as a proxy task to evaluate the pixel-level image and class-level text alignment of popular diffusion models. We conduct an in-depth analysis of pixel-text misalignment in diffusion models from the perspective of training data bias. We find that misalignment occurs in images with small sized, occluded, or rare object classes. Therefore, we propose ELBO-T2IAlign, a simple yet effective method to calibrate pixel-text alignment in diffusion models based on the evidence lower bound (ELBO) of likelihood. Our method is training-free and generic, eliminating the need to identify the specific cause of misalignment and works well across various diffusion model architectures. Extensive experiments on commonly used benchmark datasets on image segmentation and generation have verified the effectiveness of our proposed calibration approach.

Lei Liu, Zhenghao Chen, Dong Xu

3D Gaussian Splatting (3DGS) data compression is crucial for enabling efficient storage and transmission in 3D scene modeling. However, its development remains limited due to inadequate entropy models and suboptimal quantization strategies for both lossless and lossy compression scenarios, where existing methods have yet to 1) fully leverage hyperprior information to construct robust conditional entropy models, and 2) apply fine-grained, element-wise quantization strategies for improved compression granularity. In this work, we propose a novel Mixture of Priors (MoP) strategy to simultaneously address these two challenges. Specifically, inspired by the Mixture-of-Experts (MoE) paradigm, our MoP approach processes hyperprior information through multiple lightweight MLPs to generate diverse prior features, which are subsequently integrated into the MoP feature via a gating mechanism. To enhance lossless compression, the resulting MoP feature is utilized as a hyperprior to improve conditional entropy modeling. Meanwhile, for lossy compression, we employ the MoP feature as guidance information in an element-wise quantization procedure, leveraging a prior-guided Coarse-to-Fine Quantization (C2FQ) strategy with a predefined quantization step value. Specifically, we expand the quantization step value into a matrix and adaptively refine it from coarse to fine granularity, guided by the MoP feature, thereby obtaining a quantization step matrix that facilitates element-wise quantization. Extensive experiments demonstrate that our proposed 3DGS data compression framework achieves state-of-the-art performance across multiple benchmarks, including Mip-NeRF360, BungeeNeRF, DeepBlending, and Tank&Temples.

Minghan Yang, Dong Xu, Qiwen Cui et al.

In this paper, a novel second-order method called NG+ is proposed. By following the rule ``the shape of the gradient equals the shape of the parameter", we define a generalized fisher information matrix (GFIM) using the products of gradients in the matrix form rather than the traditional vectorization. Then, our generalized natural gradient direction is simply the inverse of the GFIM multiplies the gradient in the matrix form. Moreover, the GFIM and its inverse keeps the same for multiple steps so that the computational cost can be controlled and is comparable with the first-order methods. A global convergence is established under some mild conditions and a regret bound is also given for the online learning setting. Numerical results on image classification with ResNet50, quantum chemistry modeling with Schnet, neural machine translation with Transformer and recommendation system with DLRM illustrate that GN+ is competitive with the state-of-the-art methods.

Xu Dong, Binnan Zhuang, Yunxiang Mao et al.

Radars and cameras are mature, cost-effective, and robust sensors and have been widely used in the perception stack of mass-produced autonomous driving systems. Due to their complementary properties, outputs from radar detection (radar pins) and camera perception (2D bounding boxes) are usually fused to generate the best perception results. The key to successful radar-camera fusion is the accurate data association. The challenges in the radar-camera association can be attributed to the complexity of driving scenes, the noisy and sparse nature of radar measurements, and the depth ambiguity from 2D bounding boxes. Traditional rule-based association methods are susceptible to performance degradation in challenging scenarios and failure in corner cases. In this study, we propose to address radar-camera association via deep representation learning, to explore feature-level interaction and global reasoning. Additionally, we design a loss sampling mechanism and an innovative ordinal loss to overcome the difficulty of imperfect labeling and to enforce critical human-like reasoning. Despite being trained with noisy labels generated by a rule-based algorithm, our proposed method achieves a performance of 92.2% F1 score, which is 11.6% higher than the rule-based teacher. Moreover, this data-driven method also lends itself to continuous improvement via corner case mining.

Jie Liu, Chuming Li, Feng Liang et al.

As a variant of standard convolution, a dilated convolution can control effective receptive fields and handle large scale variance of objects without introducing additional computational costs. To fully explore the potential of dilated convolution, we proposed a new type of dilated convolution (referred to as inception convolution), where the convolution operations have independent dilation patterns among different axes, channels and layers. To develop a practical method for learning complex inception convolution based on the data, a simple but effective search algorithm, referred to as efficient dilation optimization (EDO), is developed. Based on statistical optimization, the EDO method operates in a low-cost manner and is extremely fast when it is applied on large scale datasets. Empirical results validate that our method achieves consistent performance gains for image recognition, object detection, instance segmentation, human detection, and human pose estimation. For instance, by simply replacing the 3x3 standard convolution in the ResNet-50 backbone with inception convolution, we significantly improve the AP of Faster R-CNN from 36.4% to 39.2% on MS COCO.

Minghan Yang, Dong Xu, Hongyu Chen et al.

In this paper, we consider stochastic second-order methods for minimizing a finite summation of nonconvex functions. One important key is to find an ingenious but cheap scheme to incorporate local curvature information. Since the true Hessian matrix is often a combination of a cheap part and an expensive part, we propose a structured stochastic quasi-Newton method by using partial Hessian information as much as possible. By further exploiting either the low-rank structure or the kronecker-product properties of the quasi-Newton approximations, the computation of the quasi-Newton direction is affordable. Global convergence to stationary point and local superlinear convergence rate are established under some mild assumptions. Numerical results on logistic regression, deep autoencoder networks and deep convolutional neural networks show that our proposed method is quite competitive to the state-of-the-art methods.

Minghan Yang, Dong Xu, Zaiwen Wen et al.

In this paper, we develop an efficient sketchy empirical natural gradient method (SENG) for large-scale deep learning problems. The empirical Fisher information matrix is usually low-rank since the sampling is only practical on a small amount of data at each iteration. Although the corresponding natural gradient direction lies in a small subspace, both the computational cost and memory requirement are still not tractable due to the high dimensionality. We design randomized techniques for different neural network structures to resolve these challenges. For layers with a reasonable dimension, sketching can be performed on a regularized least squares subproblem. Otherwise, since the gradient is a vectorization of the product between two matrices, we apply sketching on the low-rank approximations of these matrices to compute the most expensive parts. A distributed version of SENG is also developed for extremely large-scale applications. Global convergence to stationary points is established under some mild assumptions and a fast linear convergence is analyzed under the neural tangent kernel (NTK) case. Extensive experiments on convolutional neural networks show the competitiveness of SENG compared with the state-of-the-art methods. On the task ResNet50 with ImageNet-1k, SENG achieves 75.9\% Top-1 testing accuracy within 41 epochs. Experiments on the distributed large-batch training show that the scaling efficiency is quite reasonable.

Xu Dong, Pengluo Wang, Pengyue Zhang et al.

Autonomous radar has been an integral part of advanced driver assistance systems due to its robustness to adverse weather and various lighting conditions. Conventional automotive radars use digital signal processing (DSP) algorithms to process raw data into sparse radar pins that do not provide information regarding the size and orientation of the objects. In this paper, we propose a deep-learning based algorithm for radar object detection. The algorithm takes in radar data in its raw tensor representation and places probabilistic oriented bounding boxes around the detected objects in bird's-eye-view space. We created a new multimodal dataset with 102544 frames of raw radar and synchronized LiDAR data. To reduce human annotation effort we developed a scalable pipeline to automatically annotate ground truth using LiDAR as reference. Based on this dataset we developed a vehicle detection pipeline using raw radar data as the only input. Our best performing radar detection model achieves 77.28\% AP under oriented IoU of 0.3. To the best of our knowledge, this is the first attempt to investigate object detection with raw radar data for conventional corner automotive radars.

Dong Xu, Jianhui Ji, Haikuan Huang et al.

Answer selection (answer ranking) is one of the key steps in many kinds of question answering (QA) applications, where deep models have achieved state-of-the-art performance. Among these deep models, recurrent neural network (RNN) based models are most popular, typically with better performance than convolutional neural network (CNN) based models. Nevertheless, it is difficult for RNN based models to capture the information about long-range dependency among words in the sentences of questions and answers. In this paper, we propose a new deep model, called gated group self-attention (GGSA), for answer selection. GGSA is inspired by global self-attention which is originally proposed for machine translation and has not been explored in answer selection. GGSA tackles the problem of global self-attention that local and global information cannot be well distinguished. Furthermore, an interaction mechanism between questions and answers is also proposed to enhance GGSA by a residual structure. Experimental results on two popular QA datasets show that GGSA can outperform existing answer selection models to achieve state-of-the-art performance. Furthermore, GGSA can also achieve higher accuracy than global self-attention for the answer selection task, with a lower computation cost.

Dong Xu, Wu-Jun Li

Answer selection is an important subtask of question answering (QA), where deep models usually achieve better performance. Most deep models adopt question-answer interaction mechanisms, such as attention, to get vector representations for answers. When these interaction based deep models are deployed for online prediction, the representations of all answers need to be recalculated for each question. This procedure is time-consuming for deep models with complex encoders like BERT which usually have better accuracy than simple encoders. One possible solution is to store the matrix representation (encoder output) of each answer in memory to avoid recalculation. But this will bring large memory cost. In this paper, we propose a novel method, called hashing based answer selection (HAS), to tackle this problem. HAS adopts a hashing strategy to learn a binary matrix representation for each answer, which can dramatically reduce the memory cost for storing the matrix representations of answers. Hence, HAS can adopt complex encoders like BERT in the model, but the online prediction of HAS is still fast with a low memory cost. Experimental results on three popular answer selection datasets show that HAS can outperform existing models to achieve state-of-the-art performance.

Xu Dong, Swapnil Vekhande, Guohua Cao

In sparse-view Computed Tomography (CT), only a small number of projection images are taken around the object, and sinogram interpolation method has a significant impact on final image quality. When the amount of sparsity (the amount of missing views in sinogram data) is not high, conventional interpolation methods have yielded good results. When the amount of sparsity is high, more advanced sinogram interpolation methods are needed. Recently, several deep learning (DL) based sinogram interpolation methods have been proposed. However, those DL-based methods have mostly tested so far on computer simulated sinogram data rather experimentally acquired sinogram data. In this study, we developed a sinogram interpolation method for sparse-view micro-CT based on the combination of U-Net and residual learning. We applied the method to sinogram data obtained from sparse-view micro-CT experiments, where the sparsity reached 90%. The interpolated sinogram by the DL neural network was fed to FBP algorithm for reconstruction. The result shows that both RMSE and SSIM of CT image are greatly improved. The experimental results demonstrate that this sinogram interpolation method produce significantly better results over standard linear interpolation methods when the sinogram data are extremely sparse.

Tianrong Rao, Min Xu, Dong Xu

In this paper, we propose a new deep network that learns multi-level deep representations for image emotion classification (MldrNet). Image emotion can be recognized through image semantics, image aesthetics and low-level visual features from both global and local views. Existing image emotion classification works using hand-crafted features or deep features mainly focus on either low-level visual features or semantic-level image representations without taking all factors into consideration. The proposed MldrNet combines deep representations of different levels, i.e. image semantics, image aesthetics, and low-level visual features to effectively classify the emotion types of different kinds of images, such as abstract paintings and web images. Extensive experiments on both Internet images and abstract paintings demonstrate the proposed method outperforms the state-of-the-art methods using deep features or hand-crafted features. The proposed approach also outperforms the state-of-the-art methods with at least 6% performance improvement in terms of overall classification accuracy.