Yu Xue

Wing-Yin Yu, Lai-Man Po, Ray C. C. Cheung et al.

Video-based human pose transfer is a video-to-video generation task that animates a plain source human image based on a series of target human poses. Considering the difficulties in transferring highly structural patterns on the garments and discontinuous poses, existing methods often generate unsatisfactory results such as distorted textures and flickering artifacts. To address these issues, we propose a novel Deformable Motion Modulation (DMM) that utilizes geometric kernel offset with adaptive weight modulation to simultaneously perform feature alignment and style transfer. Different from normal style modulation used in style transfer, the proposed modulation mechanism adaptively reconstructs smoothed frames from style codes according to the object shape through an irregular receptive field of view. To enhance the spatio-temporal consistency, we leverage bidirectional propagation to extract the hidden motion information from a warped image sequence generated by noisy poses. The proposed feature propagation significantly enhances the motion prediction ability by forward and backward propagation. Both quantitative and qualitative experimental results demonstrate superiority over the state-of-the-arts in terms of image fidelity and visual continuity. The source code is publicly available at github.com/rocketappslab/bdmm.

Yu Xue, Haoxuan Qu, Zhuoling Li et al.

Fine-grained image classification (FGIC) has broad applications and has attracted significant research attention. In this paper, we explore a novel paradigm for solving FGIC by proposing \textbf{ToolFG}, the first tool-integrated MLLM-based framework tailored to FGIC. ToolFG enables MLLMs to autonomously and flexibly use external tools during the reasoning process, actively interact with images, and collect verifiable visual cues for distinguishing highly similar categories in a more \textit{reliable} and \textit{well-grounded} manner. To equip the model with such tool-use ability, we design a novel \textbf{MCTS-guided tool-use knowledge distillation mechanism}, which effectively mines tool-use- and FGIC-relevant knowledge from advanced proprietary MLLMs for model training. Furthermore, we propose a \textbf{model-tool co-evolution mechanism} that jointly refines the toolset and the model's tool-use policy, driving them toward a mutually adapted and FGIC-specialized state. Extensive experiments demonstrate the effectiveness of our framework.

Yu Xue, Jiafeng Qin

Differentiable neural architecture search (DARTS), as a gradient-guided search method, greatly reduces the cost of computation and speeds up the search. In DARTS, the architecture parameters are introduced to the candidate operations, but the parameters of some weight-equipped operations may not be trained well in the initial stage, which causes unfair competition between candidate operations. The weight-free operations appear in large numbers which results in the phenomenon of performance crash. Besides, a lot of memory will be occupied during training supernet which causes the memory utilization to be low. In this paper, a partial channel connection based on channel attention for differentiable neural architecture search (ADARTS) is proposed. Some channels with higher weights are selected through the attention mechanism and sent into the operation space while the other channels are directly contacted with the processed channels. Selecting a few channels with higher attention weights can better transmit important feature information into the search space and greatly improve search efficiency and memory utilization. The instability of network structure caused by random selection can also be avoided. The experimental results show that ADARTS achieved 2.46% and 17.06% classification error rates on CIFAR-10 and CIFAR-100, respectively. ADARTS can effectively solve the problem that too many skip connections appear in the search process and obtain network structures with better performance.

Yu Xue, Pengcheng Jiang, Chenchen Zhu et al.

Neural architecture search (NAS) has emerged as a powerful paradigm that enables researchers to automatically explore vast search spaces and discover efficient neural networks. However, NAS suffers from a critical bottleneck, i.e. the evaluation of numerous architectures during the search process demands substantial computing resources and time. In order to improve the efficiency of NAS, a series of methods have been proposed to reduce the evaluation time of neural architectures. However, they are not efficient enough and still only focus on the accuracy of architectures. Beyond classification accuracy, real-world applications increasingly demand more efficient and compact network architectures that balance multiple performance criteria. To address these challenges, we propose the SMEMNAS, a pairwise comparison relation-assisted multi-objective evolutionary algorithm based on a multi-population mechanism. In the SMEMNAS, a surrogate model is constructed based on pairwise comparison relations to predict the accuracy ranking of architectures, rather than the absolute accuracy. Moreover, two populations cooperate with each other in the search process, i.e. a main population that guides the evolutionary process, while a vice population that enhances search diversity. Our method aims to discover high-performance models that simultaneously optimize multiple objectives. We conduct comprehensive experiments on CIFAR-10, CIFAR-100 and ImageNet datasets to validate the effectiveness of our approach. With only a single GPU searching for 0.17 days, competitive architectures can be found by SMEMNAS which achieves 78.91% accuracy with the MAdds of 570M on the ImageNet. This work makes a significant advancement in the field of NAS.

Zhuoling Li, Hossein Rahmani, Jiarui Zhang et al.

The rapid growth of the text-to-image (T2I) community has fostered a thriving online ecosystem of expert models, which are variants of pretrained diffusion models specialized for diverse generative abilities. Yet, existing model merging methods remain limited in fully leveraging abundant online expert resources and still struggle to meet diverse in-the-wild user needs. We present DiffGraph, a novel agent-driven graph-based model merging framework, which automatically harnesses online experts and flexibly merges them for diverse user needs. Our DiffGraph constructs a scalable graph and organizes ever-expanding online experts within it through node registration and calibration. Then, DiffGraph dynamically activates specific subgraphs based on user needs, enabling flexible combinations of different experts to achieve user-desired generation. Extensive experiments show the efficacy of our method.

Farhana Yasmin, Mahade Hasan, Haipeng Liu et al.

Cardiac magnetic resonance (CMR) segmentation underpins quantitative assessment of ventricular structure and function, yet reliable delineation remains difficult due to low tissue contrast, fuzzy boundaries, and inter scan variability. We present CardiacNAS, an evolutionary neural architecture search (NAS) framework that couples a UNet like supernet with a cardiac aware search space spanning depth width, kernel size, filter size, attention, fusion, activation, dropout, and residual scaling. The search is explicitly resource aware, jointly optimizing dice similarity coefficient (DSC) and 95th percentile Hausdorff distance (HD95) versus model size and floating point operations (FLOPs) under fixed compute budgets. Candidate architectures are instantiated from the supernet, trained with proxy budgets, and evolved through crossover, mutation, and elitist selection. We evaluate on the ACDC dataset and compare against six state of the art methods, using qualitative comparisons, learning curve analyses, and design factor correlation studies. The resulting model attains 93.22% average DSC and 4.73 mm HD95 with 3.58M parameters and 14.56 GFLOPs, demonstrating a favorable accuracy efficiency trade off. Analyses indicate that searched attention and fusion choices, together with residual scaling, contribute to improved boundary fidelity and stability. CardiacNAS offers a principled, resource aware approach to deployable CMR segmentation with transparent reporting of architectural complexity and compute budgets.

Yucheng Chu, Peng He, Hang Li et al.

Short answer assessment is a vital component of science education, allowing evaluation of students' complex three-dimensional understanding. Large language models (LLMs) that possess human-like ability in linguistic tasks are increasingly popular in assisting human graders to reduce their workload. However, LLMs' limitations in domain knowledge restrict their understanding in task-specific requirements and hinder their ability to achieve satisfactory performance. Retrieval-augmented generation (RAG) emerges as a promising solution by enabling LLMs to access relevant domain-specific knowledge during assessment. In this work, we propose an adaptive RAG framework for automated grading that dynamically retrieves and incorporates domain-specific knowledge based on the question and student answer context. Our approach combines semantic search and curated educational sources to retrieve valuable reference materials. Experimental results in a science education dataset demonstrate that our system achieves an improvement in grading accuracy compared to baseline LLM approaches. The findings suggest that RAG-enhanced grading systems can serve as reliable support with efficient performance gains.

Yu Xue, Longjun Gao, Yuanqi Su et al.

Monocular Semantic Scene Completion (SSC) aims to reconstruct complete 3D semantic scenes from a single RGB image, offering a cost-effective solution for autonomous driving and robotics. However, the inherently imbalanced nature of voxel distributions, where over 93% of voxels are empty and foreground classes are rare, poses significant challenges. Existing methods often suffer from redundant emphasis on uninformative voxels and poor generalization to long-tailed categories. To address these issues, we propose VoxSAMNet (Voxel Sparsity-Aware Modulation Network), a unified framework that explicitly models voxel sparsity and semantic imbalance. Our approach introduces: (1) a Dummy Shortcut for Feature Refinement (DSFR) module that bypasses empty voxels via a shared dummy node while refining occupied ones with deformable attention; and (2) a Foreground Modulation Strategy combining Foreground Dropout (FD) and Text-Guided Image Filter (TGIF) to alleviate overfitting and enhance class-relevant features. Extensive experiments on the public benchmarks SemanticKITTI and SSCBench-KITTI-360 demonstrate that VoxSAMNet achieves state-of-the-art performance, surpassing prior monocular and stereo baselines with mIoU scores of 18.2% and 20.2%, respectively. Our results highlight the importance of sparsity-aware and semantics-guided design for efficient and accurate 3D scene completion, offering a promising direction for future research.

Haoang Lu, Yuanqi Su, Xiaoning Zhang et al.

This paper introduces VisHall3D, a novel two-stage framework for monocular semantic scene completion that aims to address the issues of feature entanglement and geometric inconsistency prevalent in existing methods. VisHall3D decomposes the scene completion task into two stages: reconstructing the visible regions (vision) and inferring the invisible regions (hallucination). In the first stage, VisFrontierNet, a visibility-aware projection module, is introduced to accurately trace the visual frontier while preserving fine-grained details. In the second stage, OcclusionMAE, a hallucination network, is employed to generate plausible geometries for the invisible regions using a noise injection mechanism. By decoupling scene completion into these two distinct stages, VisHall3D effectively mitigates feature entanglement and geometric inconsistency, leading to significantly improved reconstruction quality. The effectiveness of VisHall3D is validated through extensive experiments on two challenging benchmarks: SemanticKITTI and SSCBench-KITTI-360. VisHall3D achieves state-of-the-art performance, outperforming previous methods by a significant margin and paves the way for more accurate and reliable scene understanding in autonomous driving and other applications.

Yu Xue, Ziming Yuan

The attention mechanism is one of the most important priori knowledge to enhance convolutional neural networks. Most attention mechanisms are bound to the convolutional layer and use local or global contextual information to recalibrate the input. This is a popular attention strategy design method. Global contextual information helps the network to consider the overall distribution, while local contextual information is more general. The contextual information makes the network pay attention to the mean or maximum value of a particular receptive field. Different from the most attention mechanism, this article proposes a novel attention mechanism with the heuristic difference attention module, HDAM. HDAM's input recalibration is based on the difference between the local and global contextual information instead of the mean and maximum values. At the same time, to make different layers have a more suitable local receptive field size and increase the exibility of the local receptive field design, we use genetic algorithm to heuristically produce local receptive fields. First, HDAM extracts the mean value of the global and local receptive fields as the corresponding contextual information. Then the difference between the global and local contextual information is calculated. Finally HDAM uses this difference to recalibrate the input. In addition, we use the heuristic ability of genetic algorithm to search for the local receptive field size of each layer. Our experiments on CIFAR-10 and CIFAR-100 show that HDAM can use fewer parameters than other attention mechanisms to achieve higher accuracy. We implement HDAM with the Python library, Pytorch, and the code and models will be publicly available.

Yu Xue, Ziming Yuan, Ferrante Neri

The so-called "attention" is an efficient mechanism to improve the performance of convolutional neural networks. It uses contextual information to recalibrate the input to strengthen the propagation of informative features. However, the majority of the attention mechanisms only consider either local or global contextual information, which is singular to extract features. Moreover, many existing mechanisms directly use the contextual information to recalibrate the input, which unilaterally enhances the propagation of the informative features, but does not suppress the useless ones. This paper proposes a new attention mechanism module based on the correlation between local and global contextual information and we name this correlation as confidence. The novel attention mechanism extracts the local and global contextual information simultaneously, and calculates the confidence between them, then uses this confidence to recalibrate the input pixels. The extraction of local and global contextual information increases the diversity of features. The recalibration with confidence suppresses useless information while enhancing the informative one with fewer parameters. We use CIFAR-10 and CIFAR-100 in our experiments and explore the performance of our method's components by sufficient ablation studies. Finally, we compare our method with a various state-of-the-art convolutional neural networks and the results show that our method completely surpasses these models. We implement ConAM with the Python library, Pytorch, and the code and models will be publicly available.

Yu Xue, Ziming Yuan, Adam Slowik

With the development of automatic sleep stage classification (ASSC) techniques, many classical methods such as k-means, decision tree, and SVM have been used in automatic sleep stage classification. However, few methods explore deep learning on ASSC. Meanwhile, most deep learning methods require extensive expertise and suffer from a mass of handcrafted steps which are time-consuming especially when dealing with multi-classification tasks. In this paper, we propose an efficient five-sleep-stage classification method using convolutional neural networks (CNNs) with a novel data processing trick and we design neural architecture search (NAS) technique based on genetic algorithm (GA), NAS-G, to search for the best CNN architecture. Firstly, we attach each kernel with an adaptive coefficient to enhance the signal processing of the inputs. This can enhance the propagation of informative features and suppress the propagation of useless features in the early stage of the network. Then, we make full use of GA's heuristic search and the advantage of no need for the gradient to search for the best architecture of CNN. This can achieve a CNN with better performance than a handcrafted one in a large search space at the minimum cost. We verify the convergence of our data processing trick and compare the performance of traditional CNNs before and after using our trick. Meanwhile, we compare the performance between the CNN generated through NAS-G and the traditional CNNs with our trick. The experiments demonstrate that the convergence of CNNs with data processing trick is faster than without data processing trick and the CNN with data processing trick generated by NAS-G outperforms the handcrafted counterparts that use the data processing trick too.

Yu Xue, Yihang Tang, Xin Xu et al.

Feature selection (FS) is an important research topic in machine learning. Usually, FS is modelled as a+ bi-objective optimization problem whose objectives are: 1) classification accuracy; 2) number of features. One of the main issues in real-world applications is missing data. Databases with missing data are likely to be unreliable. Thus, FS performed on a data set missing some data is also unreliable. In order to directly control this issue plaguing the field, we propose in this study a novel modelling of FS: we include reliability as the third objective of the problem. In order to address the modified problem, we propose the application of the non-dominated sorting genetic algorithm-III (NSGA-III). We selected six incomplete data sets from the University of California Irvine (UCI) machine learning repository. We used the mean imputation method to deal with the missing data. In the experiments, k-nearest neighbors (K-NN) is used as the classifier to evaluate the feature subsets. Experimental results show that the proposed three-objective model coupled with NSGA-III efficiently addresses the FS problem for the six data sets included in this study.