Pu Wang

Ekkasit Pinyoanuntapong, Ayman Ali, Pu Wang et al.

Most existing gait recognition methods are appearance-based, which rely on the silhouettes extracted from the video data of human walking activities. The less-investigated skeleton-based gait recognition methods directly learn the gait dynamics from 2D/3D human skeleton sequences, which are theoretically more robust solutions in the presence of appearance changes caused by clothes, hairstyles, and carrying objects. However, the performance of skeleton-based solutions is still largely behind the appearance-based ones. This paper aims to close such performance gap by proposing a novel network model, GaitMixer, to learn more discriminative gait representation from skeleton sequence data. In particular, GaitMixer follows a heterogeneous multi-axial mixer architecture, which exploits the spatial self-attention mixer followed by the temporal large-kernel convolution mixer to learn rich multi-frequency signals in the gait feature maps. Experiments on the widely used gait database, CASIA-B, demonstrate that GaitMixer outperforms the previous SOTA skeleton-based methods by a large margin while achieving a competitive performance compared with the representative appearance-based solutions. Code will be available at https://github.com/exitudio/gaitmixer

Pu Wang, Hugo Van hamme

End-to-end spoken language understanding (SLU) systems benefit from pretraining on large corpora, followed by fine-tuning on application-specific data. The resulting models are too large for on-edge applications. For instance, BERT-based systems contain over 110M parameters. Observing the model is overparameterized, we propose lean transformer structure where the dimension of the attention mechanism is automatically reduced using group sparsity. We propose a variant where the learned attention subspace is transferred to an attention bottleneck layer. In a low-resource setting and without pre-training, the resulting compact SLU model achieves accuracies competitive with pre-trained large models.

Ekkasit Pinyoanuntapong, Ayman Ali, Kalvik Jakkala et al.

mmWave radar-based gait recognition is a novel user identification method that captures human gait biometrics from mmWave radar return signals. This technology offers privacy protection and is resilient to weather and lighting conditions. However, its generalization performance is yet unknown and limits its practical deployment. To address this problem, in this paper, a non-synthetic dataset is collected and analyzed to reveal the presence of spatial and temporal domain shifts in mmWave gait biometric data, which significantly impacts identification accuracy. To mitigate this issue, a novel self-aligned domain adaptation method called GaitSADA is proposed. GaitSADA improves system generalization performance by using a two-stage semi-supervised model training approach. The first stage employs semi-supervised contrastive learning to learn a compact gait representation from both source and target domain data, aligning source-target domain distributions implicitly. The second stage uses semi-supervised consistency training with centroid alignment to further close source-target domain gap by pseudo-labelling the target-domain samples, clustering together the samples belonging to the same class but from different domains, and pushing the class centroid close to the weight vector of each class. Experiments show that GaitSADA outperforms representative domain adaptation methods with an improvement ranging from 15.41\% to 26.32\% on average accuracy in low data regimes. Code and dataset will be available at https://exitudio.github.io/GaitSADA

Ayman Ali, Ekkasit Pinyoanuntapong, Pu Wang et al.

Recently, there has been a remarkable increase in the interest towards skeleton-based action recognition within the research community, owing to its various advantageous features, including computational efficiency, representative features, and illumination invariance. Despite this, researchers continue to explore and investigate the most optimal way to represent human actions through skeleton representation and the extracted features. As a result, the growth and availability of human action recognition datasets have risen substantially. In addition, deep learning-based algorithms have gained widespread popularity due to the remarkable advancements in various computer vision tasks. Most state-of-the-art contributions in skeleton-based action recognition incorporate a Graph Neural Network (GCN) architecture for representing the human body and extracting features. Our research demonstrates that Convolutional Neural Networks (CNNs) can attain comparable results to GCN, provided that the proper training techniques, augmentations, and optimizers are applied. Our approach has been rigorously validated, and we have achieved a score of 95% on the NTU-60 dataset

Ayman Ali, Ekkasit Pinyoanuntapong, Pu Wang et al.

In this research, we address the challenge faced by existing deep learning-based human mesh reconstruction methods in balancing accuracy and computational efficiency. These methods typically prioritize accuracy, resulting in large network sizes and excessive computational complexity, which may hinder their practical application in real-world scenarios, such as virtual reality systems. To address this issue, we introduce a modular multi-stage lightweight graph-based transformer network for human pose and shape estimation from 2D human pose, a pose-based human mesh reconstruction approach that prioritizes computational efficiency without sacrificing reconstruction accuracy. Our method consists of a 2D-to-3D lifter module that utilizes graph transformers to analyze structured and implicit joint correlations in 2D human poses, and a mesh regression module that combines the extracted pose features with a mesh template to produce the final human mesh parameters.

Qucheng Peng, Ce Zheng, Zhengming Ding et al.

Human pose estimation (HPE) has received increasing attention recently due to its wide application in motion analysis, virtual reality, healthcare, etc. However, it suffers from the lack of labeled diverse real-world datasets due to the time- and labor-intensive annotation. To cope with the label deficiency issue, one common solution is to train the HPE models with easily available synthetic datasets (source) and apply them to real-world data (target) through domain adaptation (DA). Unfortunately, prevailing domain adaptation techniques within the HPE domain remain predominantly fixated on effecting alignment and aggregation between source and target features, often sidestepping the crucial task of excluding domain-specific representations. To rectify this, we introduce a novel framework that capitalizes on both representation aggregation and segregation for domain adaptive human pose estimation. Within this framework, we address the network architecture aspect by disentangling representations into distinct domain-invariant and domain-specific components, facilitating aggregation of domain-invariant features while simultaneously segregating domain-specific ones. Moreover, we tackle the discrepancy measurement facet by delving into various keypoint relationships and applying separate aggregation or segregation mechanisms to enhance alignment. Extensive experiments on various benchmarks, e.g., Human3.6M, LSP, H3D, and FreiHand, show that our method consistently achieves state-of-the-art performance. The project is available at \url{https://github.com/davidpengucf/EPIC}.

Xingchi Chen, Pu Wang, Xuerui Li et al.

Ultra-High-Definition (UHD) image dehazing faces challenges such as limited scene adaptability in prior-based methods and high computational complexity with color distortion in deep learning approaches. To address these issues, we propose 4KDehazeFlow, a novel method based on Flow Matching and the Haze-Aware vector field. This method models the dehazing process as a progressive optimization of continuous vector field flow, providing efficient data-driven adaptive nonlinear color transformation for high-quality dehazing. Specifically, our method has the following advantages: 1) 4KDehazeFlow is a general method compatible with various deep learning networks, without relying on any specific network architecture. 2) We propose a learnable 3D lookup table (LUT) that encodes haze transformation parameters into a compact 3D mapping matrix, enabling efficient inference through precomputed mappings. 3) We utilize a fourth-order Runge-Kutta (RK4) ordinary differential equation (ODE) solver to stably solve the dehazing flow field through an accurate step-by-step iterative method, effectively suppressing artifacts. Extensive experiments show that 4KDehazeFlow exceeds seven state-of-the-art methods. It delivers a 2dB PSNR increase and better performance in dense haze and color fidelity.

Matias Mendieta, Taojiannan Yang, Pu Wang et al.

Federated learning (FL) is a promising strategy for performing privacy-preserving, distributed learning with a network of clients (i.e., edge devices). However, the data distribution among clients is often non-IID in nature, making efficient optimization difficult. To alleviate this issue, many FL algorithms focus on mitigating the effects of data heterogeneity across clients by introducing a variety of proximal terms, some incurring considerable compute and/or memory overheads, to restrain local updates with respect to the global model. Instead, we consider rethinking solutions to data heterogeneity in FL with a focus on local learning generality rather than proximal restriction. To this end, we first present a systematic study informed by second-order indicators to better understand algorithm effectiveness in FL. Interestingly, we find that standard regularization methods are surprisingly strong performers in mitigating data heterogeneity effects. Based on our findings, we further propose a simple and effective method, FedAlign, to overcome data heterogeneity and the pitfalls of previous methods. FedAlign achieves competitive accuracy with state-of-the-art FL methods across a variety of settings while minimizing computation and memory overhead. Code is available at https://github.com/mmendiet/FedAlign

Taojiannan Yang, Sijie Zhu, Matias Mendieta et al.

Most existing deep neural networks are static, which means they can only do inference at a fixed complexity. But the resource budget can vary substantially across different devices. Even on a single device, the affordable budget can change with different scenarios, and repeatedly training networks for each required budget would be incredibly expensive. Therefore, in this work, we propose a general method called MutualNet to train a single network that can run at a diverse set of resource constraints. Our method trains a cohort of model configurations with various network widths and input resolutions. This mutual learning scheme not only allows the model to run at different width-resolution configurations but also transfers the unique knowledge among these configurations, helping the model to learn stronger representations overall. MutualNet is a general training methodology that can be applied to various network structures (e.g., 2D networks: MobileNets, ResNet, 3D networks: SlowFast, X3D) and various tasks (e.g., image classification, object detection, segmentation, and action recognition), and is demonstrated to achieve consistent improvements on a variety of datasets. Since we only train the model once, it also greatly reduces the training cost compared to independently training several models. Surprisingly, MutualNet can also be used to significantly boost the performance of a single network, if dynamic resource constraint is not a concern. In summary, MutualNet is a unified method for both static and adaptive, 2D and 3D networks. Codes and pre-trained models are available at \url{https://github.com/taoyang1122/MutualNet}.

Ekkasit Pinyoanuntapong, Muhammad Usama Saleem, Pu Wang et al.

Generating human motion from text has been dominated by denoising motion models either through diffusion or generative masking process. However, these models face great limitations in usability by requiring prior knowledge of the motion length. Conversely, autoregressive motion models address this limitation by adaptively predicting motion endpoints, at the cost of degraded generation quality and editing capabilities. To address these challenges, we propose Bidirectional Autoregressive Motion Model (BAMM), a novel text-to-motion generation framework. BAMM consists of two key components: (1) a motion tokenizer that transforms 3D human motion into discrete tokens in latent space, and (2) a masked self-attention transformer that autoregressively predicts randomly masked tokens via a hybrid attention masking strategy. By unifying generative masked modeling and autoregressive modeling, BAMM captures rich and bidirectional dependencies among motion tokens, while learning the probabilistic mapping from textual inputs to motion outputs with dynamically-adjusted motion sequence length. This feature enables BAMM to simultaneously achieving high-quality motion generation with enhanced usability and built-in motion editability. Extensive experiments on HumanML3D and KIT-ML datasets demonstrate that BAMM surpasses current state-of-the-art methods in both qualitative and quantitative measures. Our project page is available at https://exitudio.github.io/BAMM-page

Pu Wang, Zhihua Zhang, Dianjie Lu et al.

Since human and environmental factors interfere, captured polyp images usually suffer from issues such as dim lighting, blur, and overexposure, which pose challenges for downstream polyp segmentation tasks. To address the challenges of noise-induced degradation in polyp images, we present AgentPolyp, a novel framework integrating CLIP-based semantic guidance and dynamic image enhancement with a lightweight neural network for segmentation. The agent first evaluates image quality using CLIP-driven semantic analysis (e.g., identifying ``low-contrast polyps with vascular textures") and adapts reinforcement learning strategies to dynamically apply multi-modal enhancement operations (e.g., denoising, contrast adjustment). A quality assessment feedback loop optimizes pixel-level enhancement and segmentation focus in a collaborative manner, ensuring robust preprocessing before neural network segmentation. This modular architecture supports plug-and-play extensions for various enhancement algorithms and segmentation networks, meeting deployment requirements for endoscopic devices.

Muhammad Usama Saleem, Ekkasit Pinyoanuntapong, Pu Wang et al.

Human mesh recovery (HMR) is crucial in many computer vision applications; from health to arts and entertainment. HMR from monocular images has predominantly been addressed by deterministic methods that output a single prediction for a given 2D image. However, HMR from a single image is an ill-posed problem due to depth ambiguity and occlusions. Probabilistic methods have attempted to address this by generating and fusing multiple plausible 3D reconstructions, but their performance has often lagged behind deterministic approaches. In this paper, we introduce GenHMR, a novel generative framework that reformulates monocular HMR as an image-conditioned generative task, explicitly modeling and mitigating uncertainties in the 2D-to-3D mapping process. GenHMR comprises two key components: (1) a pose tokenizer to convert 3D human poses into a sequence of discrete tokens in a latent space, and (2) an image-conditional masked transformer to learn the probabilistic distributions of the pose tokens, conditioned on the input image prompt along with randomly masked token sequence. During inference, the model samples from the learned conditional distribution to iteratively decode high-confidence pose tokens, thereby reducing 3D reconstruction uncertainties. To further refine the reconstruction, a 2D pose-guided refinement technique is proposed to directly fine-tune the decoded pose tokens in the latent space, which forces the projected 3D body mesh to align with the 2D pose clues. Experiments on benchmark datasets demonstrate that GenHMR significantly outperforms state-of-the-art methods. Project website can be found at https://m-usamasaleem.github.io/publication/GenHMR/GenHMR.html

Ce Zheng, Matias Mendieta, Pu Wang et al.

Existing deep learning-based human mesh reconstruction approaches have a tendency to build larger networks in order to achieve higher accuracy. Computational complexity and model size are often neglected, despite being key characteristics for practical use of human mesh reconstruction models (e.g. virtual try-on systems). In this paper, we present GTRS, a lightweight pose-based method that can reconstruct human mesh from 2D human pose. We propose a pose analysis module that uses graph transformers to exploit structured and implicit joint correlations, and a mesh regression module that combines the extracted pose feature with the mesh template to reconstruct the final human mesh. We demonstrate the efficiency and generalization of GTRS by extensive evaluations on the Human3.6M and 3DPW datasets. In particular, GTRS achieves better accuracy than the SOTA pose-based method Pose2Mesh while only using 10.2% of the parameters (Params) and 2.5% of the FLOPs on the challenging in-the-wild 3DPW dataset. Code will be publicly available.

Pinyarash Pinyoanuntapong, Tagore Pothuneedi, Ravikumar Balakrishnan et al.

Federated Learning (FL) over wireless multi-hop edge computing networks, i.e., multi-hop FL, is a cost-effective distributed on-device deep learning paradigm. This paper presents FedEdge simulator, a high-fidelity Linux-based simulator, which enables fast prototyping, sim-to-real code, and knowledge transfer for multi-hop FL systems. FedEdge simulator is built on top of the hardware-oriented FedEdge experimental framework with a new extension of the realistic physical layer emulator. This emulator exploits trace-based channel modeling and dynamic link scheduling to minimize the reality gap between the simulator and the physical testbed. Our initial experiments demonstrate the high fidelity of the FedEdge simulator and its superior performance on sim-to-real knowledge transfer in reinforcement learning-optimized multi-hop FL.

Pinyarash Pinyoanuntapong, Prabhu Janakaraj, Ravikumar Balakrishnan et al.

Federated learning (FL) is a distributed machine learning technology for next-generation AI systems that allows a number of workers, i.e., edge devices, collaboratively learn a shared global model while keeping their data locally to prevent privacy leakage. Enabling FL over wireless multi-hop networks can democratize AI and make it accessible in a cost-effective manner. However, the noisy bandwidth-limited multi-hop wireless connections can lead to delayed and nomadic model updates, which significantly slows down the FL convergence speed. To address such challenges, this paper aims to accelerate FL convergence over wireless edge by optimizing the multi-hop federated networking performance. In particular, the FL convergence optimization problem is formulated as a Markov decision process (MDP). To solve such MDP, multi-agent reinforcement learning (MA-RL) algorithms along with domain-specific action space refining schemes are developed, which online learn the delay-minimum forwarding paths to minimize the model exchange latency between the edge devices (i.e., workers) and the remote server. To validate the proposed solutions, FedEdge is developed and implemented, which is the first experimental framework in the literature for FL over multi-hop wireless edge computing networks. FedEdge allows us to fast prototype, deploy, and evaluate novel FL algorithms along with RL-based system optimization methods in real wireless devices. Moreover, a physical experimental testbed is implemented by customizing the widely adopted Linux wireless routers and ML computing nodes.Finally, our experimentation results on the testbed show that the proposed network-accelerated FL system can practically and significantly improve FL convergence speed, compared to the FL system empowered by the production-grade commercially available wireless networking protocol, BATMAN-Adv.

Zhi Sun, Sarankumar Balakrishnan, Lu Su et al.

With the wide bandwidths in millimeter wave (mmWave) frequency band that results in unprecedented accuracy, mmWave sensing has become vital for many applications, especially in autonomous vehicles (AVs). In addition, mmWave sensing has superior reliability compared to other sensing counterparts such as camera and LiDAR, which is essential for safety-critical driving. Therefore, it is critical to understand the security vulnerabilities and improve the security and reliability of mmWave sensing in AVs. To this end, we perform the end-to-end security analysis of a mmWave-based sensing system in AVs, by designing and implementing practical physical layer attack and defense strategies in a state-of-the-art mmWave testbed and an AV testbed in real-world settings. Various strategies are developed to take control of the victim AV by spoofing its mmWave sensing module, including adding fake obstacles at arbitrary locations and faking the locations of existing obstacles. Five real-world attack scenarios are constructed to spoof the victim AV and force it to make dangerous driving decisions leading to a fatal crash. Field experiments are conducted to study the impact of the various attack scenarios using a Lincoln MKZ-based AV testbed, which validate that the attacker can indeed assume control of the victim AV to compromise its security and safety. To defend the attacks, we design and implement a challenge-response authentication scheme and a RF fingerprinting scheme to reliably detect aforementioned spoofing attacks.

Sarankumar Balakrishnan, Shreya Gupta, Arupjyoti Bhuyan et al.

With millimeter wave (mmWave) wireless communication envisioned to be the key enabler of next generation high data rate wireless networks, security is of paramount importance. While conventional security measures in wireless networks operate at a higher layer of the protocol stack, physical layer security utilizes unique device dependent hardware features to identify and authenticate legitimate devices. In this work, we identify that the manufacturing tolerances in the antenna arrays used in mmWave devices contribute to a beam pattern that is unique to each device, and to that end we propose a novel device fingerprinting scheme based on the unique beam patterns used by the mmWave devices. Specifically, we propose a fingerprinting scheme with multiple access points (APs) to take advantage of the rich spatial-temporal information of the beam pattern. We perform comprehensive experiments with commercial off-the-shelf mmWave devices to validate the reliability performance of our proposed method under various scenarios. We also compare our beam pattern feature with a conventional physical layer feature namely power spectral density feature (PSD). To that end, we implement PSD feature based fingerprinting for mmWave devices. We show that the proposed multiple APs scheme is able to achieve over 99% identification accuracy for stationary LOS and NLOS scenarios and significantly outperform the PSD based method. For mobility scenarios, the overall identification accuracy is 96%. In addition, we perform security analysis of our proposed beam pattern fingerprinting system and PSD fingerprinting system by studying the feasibility of performing impersonation attacks. We design and implement an impersonation attack mechanism for mmWave wireless networks using state-of-the-art 60 GHz software defined radios. We discuss our findings and their implications on the security of the mmWave wireless networks.

Kalvik Jakkala, Arupjyoti Bhuya, Zhi Sun et al.

Gait is a person's natural walking style and a complex biological process that is unique to each person. Recently, the channel state information (CSI) of WiFi devices have been exploited to capture human gait biometrics for user identification. However, the performance of existing CSI-based gait identification systems is far from satisfactory. They can only achieve limited identification accuracy (maximum $93\%$) only for a very small group of people (i.e., between 2 to 10). To address such challenge, an end-to-end deep CSI learning system is developed, which exploits deep neural networks to automatically learn the salient gait features in CSI data that are discriminative enough to distinguish different people Firstly, the raw CSI data are sanitized through window-based denoising, mean centering and normalization. The sanitized data is then passed to a residual deep convolutional neural network (DCNN), which automatically extracts the hierarchical features of gait-signatures embedded in the CSI data. Finally, a softmax classifier utilizes the extracted features to make the final prediction about the identity of the user. In a typical indoor environment, a top-1 accuracy of $97.12 \pm 1.13\%$ is achieved for a dataset of 30 people.

Pu Wang, Michael Emmerich, Rui Li et al.

ROC is usually used to analyze the performance of classifiers in data mining. ROC convex hull (ROCCH) is the least convex major-ant (LCM) of the empirical ROC curve, and covers potential optima for the given set of classifiers. Generally, ROC performance maximization could be considered to maximize the ROCCH, which also means to maximize the true positive rate (tpr) and minimize the false positive rate (fpr) for each classifier in the ROC space. However, tpr and fpr are conflicting with each other in the ROCCH optimization process. Though ROCCH maximization problem seems like a multi-objective optimization problem (MOP), the special characters make it different from traditional MOP. In this work, we will discuss the difference between them and propose convex hull-based multi-objective genetic programming (CH-MOGP) to solve ROCCH maximization problems. Convex hull-based sort is an indicator based selection scheme that aims to maximize the area under convex hull, which serves as a unary indicator for the performance of a set of points. A selection procedure is described that can be efficiently implemented and follows similar design principles than classical hyper-volume based optimization algorithms. It is hypothesized that by using a tailored indicator-based selection scheme CH-MOGP gets more efficient for ROC convex hull approximation than algorithms which compute all Pareto optimal points. To test our hypothesis we compare the new CH-MOGP to MOGP with classical selection schemes, including NSGA-II, MOEA/D) and SMS-EMOA. Meanwhile, CH-MOGP is also compared with traditional machine learning algorithms such as C4.5, Naive Bayes and Prie. Experimental results based on 22 well-known UCI data sets show that CH-MOGP outperforms significantly traditional EMOAs.