Yongjie Hou

Longfei Liu, Yongjie Hou, Yang Li et al.

Deploying high-performance dense prediction models on resource-constrained edge devices remains challenging due to strict limits on computation and memory. In practice, lightweight systems for object detection, instance segmentation, and pose estimation are still dominated by CNN-based architectures such as YOLO, while compact Vision Transformers (ViTs) often struggle to achieve similarly strong accuracy efficiency tradeoff, even with large scale pretraining. We argue that this gap is largely due to insufficient task specific representation learning in small scale ViTs, rather than an inherent mismatch between ViTs and edge dense prediction. To address this issue, we introduce EdgeCrafter, a unified compact ViT framework for edge dense prediction centered on ECDet, a detection model built from a distilled compact backbone and an edge-friendly encoder decoder design. On the COCO dataset, ECDet-S achieves 51.7 AP with fewer than 10M parameters using only COCO annotations. For instance segmentation, ECInsSeg achieves performance comparable to RF-DETR while using substantially fewer parameters. For pose estimation, ECPose-X reaches 74.8 AP, significantly outperforming YOLO26Pose-X (71.6 AP) despite the latter's reliance on extensive Objects365 pretraining. These results show that compact ViTs, when paired with task-specialized distillation and edge-aware design, can be a practical and competitive option for edge dense prediction. Code is available at: https://intellindust-ai-lab.github.io/projects/EdgeCrafter/

Yunpeng Gong, Yongjie Hou, Zhenzhong Wang et al.

Neural network solvers for partial differential equations (PDEs) have made significant progress, yet they continue to face challenges related to data scarcity and model robustness. Traditional data augmentation methods, which leverage symmetry or invariance, impose strong assumptions on physical systems that often do not hold in dynamic and complex real-world applications. To address this research gap, this study introduces a universal learning strategy for neural network PDEs, named Systematic Model Augmentation for Robust Training (SMART). By focusing on challenging and improving the model's weaknesses, SMART reduces generalization error during training under data-scarce conditions, leading to significant improvements in prediction accuracy across various PDE scenarios. The effectiveness of the proposed method is demonstrated through both theoretical analysis and extensive experimentation. The code will be available.

Jiajing Lin, Zhenzhong Wang, Yongjie Hou et al.

4D content generation focuses on creating dynamic 3D objects that change over time. Existing methods primarily rely on pre-trained video diffusion models, utilizing sampling processes or reference videos. However, these approaches face significant challenges. Firstly, the generated 4D content often fails to adhere to real-world physics since video diffusion models do not incorporate physical priors. Secondly, the extensive sampling process and the large number of parameters in diffusion models result in exceedingly time-consuming generation processes. To address these issues, we introduce Phy124, a novel, fast, and physics-driven method for controllable 4D content generation from a single image. Phy124 integrates physical simulation directly into the 4D generation process, ensuring that the resulting 4D content adheres to natural physical laws. Phy124 also eliminates the use of diffusion models during the 4D dynamics generation phase, significantly speeding up the process. Phy124 allows for the control of 4D dynamics, including movement speed and direction, by manipulating external forces. Extensive experiments demonstrate that Phy124 generates high-fidelity 4D content with significantly reduced inference times, achieving stateof-the-art performance. The code and generated 4D content are available at the provided link: https://anonymous.4open.science/r/BBF2/.

Yunpeng Gong, Yongjie Hou, Chuangliang Zhang et al.

Person Re-identification (re-ID) in computer vision aims to recognize and track individuals across different cameras. While previous research has mainly focused on challenges like pose variations and lighting changes, the impact of extreme capture conditions is often not adequately addressed. These extreme conditions, including varied lighting, camera styles, angles, and image distortions, can significantly affect data distribution and re-ID accuracy. Current research typically improves model generalization under normal shooting conditions through data augmentation techniques such as adjusting brightness and contrast. However, these methods pay less attention to the robustness of models under extreme shooting conditions. To tackle this, we propose a multi-mode synchronization learning (MMSL) strategy . This approach involves dividing images into grids, randomly selecting grid blocks, and applying data augmentation methods like contrast and brightness adjustments. This process introduces diverse transformations without altering the original image structure, helping the model adapt to extreme variations. This method improves the model's generalization under extreme conditions and enables learning diverse features, thus better addressing the challenges in re-ID. Extensive experiments on a simulated test set under extreme conditions have demonstrated the effectiveness of our method. This approach is crucial for enhancing model robustness and adaptability in real-world scenarios, supporting the future development of person re-identification technology.

Yunpeng Gong, Chuangliang Zhang, Yongjie Hou et al.

In the contemporary of deep learning, where models often grapple with the challenge of simultaneously achieving robustness against adversarial attacks and strong generalization capabilities, this study introduces an innovative Local Feature Masking (LFM) strategy aimed at fortifying the performance of Convolutional Neural Networks (CNNs) on both fronts. During the training phase, we strategically incorporate random feature masking in the shallow layers of CNNs, effectively alleviating overfitting issues, thereby enhancing the model's generalization ability and bolstering its resilience to adversarial attacks. LFM compels the network to adapt by leveraging remaining features to compensate for the absence of certain semantic features, nurturing a more elastic feature learning mechanism. The efficacy of LFM is substantiated through a series of quantitative and qualitative assessments, collectively showcasing a consistent and significant improvement in CNN's generalization ability and resistance against adversarial attacks--a phenomenon not observed in current and prior methodologies. The seamless integration of LFM into established CNN frameworks underscores its potential to advance both generalization and adversarial robustness within the deep learning paradigm. Through comprehensive experiments, including robust person re-identification baseline generalization experiments and adversarial attack experiments, we demonstrate the substantial enhancements offered by LFM in addressing the aforementioned challenges. This contribution represents a noteworthy stride in advancing robust neural network architectures.

Shihua Huang, Yongjie Hou, Longfei Liu et al.

Benefiting from the simplicity and effectiveness of Dense O2O and MAL, DEIM has become the mainstream training framework for real-time DETRs, significantly outperforming the YOLO series. In this work, we extend it with DINOv3 features, resulting in DEIMv2. DEIMv2 spans eight model sizes from X to Atto, covering GPU, edge, and mobile deployment. For the X, L, M, and S variants, we adopt DINOv3-pretrained or distilled backbones and introduce a Spatial Tuning Adapter (STA), which efficiently converts DINOv3's single-scale output into multi-scale features and complements strong semantics with fine-grained details to enhance detection. For ultra-lightweight models (Nano, Pico, Femto, and Atto), we employ HGNetv2 with depth and width pruning to meet strict resource budgets. Together with a simplified decoder and an upgraded Dense O2O, this unified design enables DEIMv2 to achieve a superior performance-cost trade-off across diverse scenarios, establishing new state-of-the-art results. Notably, our largest model, DEIMv2-X, achieves 57.8 AP with only 50.3 million parameters, surpassing prior X-scale models that require over 60 million parameters for just 56.5 AP. On the compact side, DEIMv2-S is the first sub-10 million model (9.71 million) to exceed the 50 AP milestone on COCO, reaching 50.9 AP. Even the ultra-lightweight DEIMv2-Pico, with just 1.5 million parameters, delivers 38.5 AP, matching YOLOv10-Nano (2.3 million) with around 50 percent fewer parameters. Our code and pre-trained models are available at https://github.com/Intellindust-AI-Lab/DEIMv2

Yuzhou Tang, Dejun Xu, Yongjie Hou et al.

Photorealistic 3D reconstruction of unstructured real-world scenes remains challenging due to complex illumination variations and transient occlusions. Existing methods based on Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) struggle with inefficient light decoupling and structure-agnostic occlusion handling. To address these limitations, we propose NexusSplats, an approach tailored for efficient and high-fidelity 3D scene reconstruction under complex lighting and occlusion conditions. In particular, NexusSplats leverages a hierarchical light decoupling strategy that performs centralized appearance learning, efficiently and effectively decoupling varying lighting conditions. Furthermore, a structure-aware occlusion handling mechanism is developed, establishing a nexus between 3D and 2D structures for fine-grained occlusion handling. Experimental results demonstrate that NexusSplats achieves state-of-the-art rendering quality and reduces the number of total parameters by 65.4\%, leading to 2.7$\times$ faster reconstruction.

Yunpeng Gong, Yongjie Hou, Jiangming Shi et al.

Sketch based person re-identification aims to match hand-drawn sketches with RGB surveillance images, but remains challenging due to significant modality gaps and limited annotated data. To address this, we introduce KTCAA, a theoretically grounded framework for few-shot cross-modal generalization. Motivated by generalization theory, we identify two key factors influencing target domain risk: (1) domain discrepancy, which quantifies the alignment difficulty between source and target distributions; and (2) perturbation invariance, which evaluates the model's robustness to modality shifts. Based on these insights, we propose two components: (1) Alignment Augmentation (AA), which applies localized sketch-style transformations to simulate target distributions and facilitate progressive alignment; and (2) Knowledge Transfer Catalyst (KTC), which enhances invariance by introducing worst-case perturbations and enforcing consistency. These modules are jointly optimized under a meta-learning paradigm that transfers alignment knowledge from data-rich RGB domains to sketch-based scenarios. Experiments on multiple benchmarks demonstrate that KTCAA achieves state-of-the-art performance, particularly in data-scarce conditions.