HYHuimin Yu

Huimin Yu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Huimin Yu has a total of 61 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    VSF: A Multimodal Few-Shot Plant Disease Recognition Method in Agricultural Scenarios

    by Houkui Zhou, Xinzhan Li, Yuan Nie, Huimin Yu, Haoji Hu, Lujun Lin

    Identifying plant diseases presents challenges like small sample sizes, imbalanced classes, and intricate background noise, which constrain the generalizability and resilience of conventional deep learning methods that rely on single-modal images. To overcome these limitations, this study introduces a novel approach for recognizing plant diseases with limited examples, utilizing multiple modalities that combine visual and textual data to enhance the model’s discriminative power and generalization performance. First, the official definition of plant diseases is semantically enriched using large language models to automatically generate high-quality textual descriptions that include symptoms, affected plant parts, and agricultural context characteristics. Second, a visual–semantic multimodal fusion framework is developed, incorporating a Two-Stage Modality Fusion mechanism that projects semantic and visual features into a unified feature space to optimize cross-modal feature interaction. The experiments were performed on the PlantVillage dataset and the PlantDoc dataset to validate the N-way K-shot task. Results show that the proposed method achieved accuracies of 79.01% and 90.51% under the 5-way 1-shot and 5-way 5-shot tasks on the PlantVillage dataset, and 59.82% and 70.61% on the PlantDoc dataset. These results outperformed traditional unimodal and existing multimodal approaches. This paper provides an efficient and widely applicable multimodal approach to few-shot plant disease identification in agriculture.

  • Structured Distilled 3D Gait Fields for Clothes-Changing Person Re-Identification in Videos

    by Yubin Wang, Huimin Yu, Haoji Hu, Yongsheng Gao

    3D gait representation, intrinsically tied to identity, is of great importance in tackling the challenge of video-based clothes-changing person re-identification (VCCRe-ID), where individuals’ clothing appearances vary across video clips captured over an extended duration. Most existing methods derive 3D gait features from sequential 3D skeletons or parameters of a predefined 3D human model (e.g. SMPL). However, the sparsity of 3D skeletons and the low-dimensional nature of 3D human model parameters not only limit their expressive power and discriminative ability, but also impede alignment with higher-dimensional appearance features. To address these issues, we propose Structured Distilled 3D Gait Fields (SD-3DGF) to learn better aligned, more expressive, and discriminative 3D gait features for VCCRe-ID. SD-3DGF defines distilled feature fields that evolve over time, allowing the simulation of 3D gait dynamics from temporal 2D descriptors. Specifically, it densely transfers frame-level shape knowledge from structured 3D space (i.e., 3D mesh surface) to the 2D image domain via 3D-2D knowledge distillation, and models local spatiotemporal correlations among the distilled features to extract 3D gait features. This learning paradigm makes 3D gait features more compatible with appearance features, as they are naturally aligned in the spatiotemporal domain. To fully exploit this potential, we develop a novel Modality-Representative Interactive Encoding (MRIE) module, which facilitates efficient gait-appearance interaction and joint encoding via global and local representative tokens. Experiments on both VCCRe-ID and standard video-based Re-ID benchmarks demonstrate the effectiveness of our method. The code is available at https://github.com/yubinwang2021/SD-3DGF.

  • Open Access

    HGCS-Det: A Deep Learning-Based Solution for Localizing and Recognizing Household Garbage in Complex Scenarios

    by Houkui Zhou, Chang Cheng, Zhongyi Xia, Qifeng Ding, Qinqin Liao, Qun Wang, Huimin Yu, Haoji Hu, Guangqun Zhang, Junguo Hu, et al.

    With the rise of deep learning technology, intelligent garbage detection provides a new idea for garbage classification management. However, due to the interference of complex environments, coupled with the influence of the irregular features of garbage, garbage detection in complex scenarios still faces significant challenges. Moreover, some of the existing research suffer from shortcomings in either their precision or real-time performance, particularly when applied to complex garbage detection scenarios. Therefore, this paper proposes a model based on YOLOv8, namely HGCS-Det, for detecting garbage in complex scenarios. The HGCS-Det model is designed as follows: Firstly, the normalization attention module is introduced to calibrate the model’s attention to targets and to suppress the environmental noise interference information. Additionally, to weigh the attention-feature contributions, an Attention Feature Fusion module is employed to complement the attention weights of each channel. Subsequently, an Instance Boundary Reinforcement module is established to capture the fine-grained features of garbage by combining strong gradient information with semantic information. Finally, the Slide Loss function is applied to dynamically weight hard samples arising from the complex detection environments to improve the recognition accuracy of hard samples. With only a slight increase in parameters (3.02M), HGCS-Det achieves a 93.6% mean average precision (mAP) and 86 FPS on the public HGI30 dataset, which is a 3.33% higher mAP value than from YOLOv12, and outperforms the state-of-the-art (SOTA) methods in both efficiency and applicability. Notably, HGCS-Det maintains a lightweight architecture while enhancing the detection accuracy, enabling real-time performance even in resource-constrained environments. These characteristics significantly improve its practical applicability, making the model well suited for deployment in embedded devices and real-world garbage classification systems. This method can serve as a valuable technical reference for the engineering application of garbage classification.


  • Open Access

    Optimizing Backbone Networks Through Hybrid–Modal Fusion: A New Strategy for Waste Classification

    by Houkui Zhou, Qifeng Ding, Chang Cheng, Qinqin Liao, Qun Wang, Huimin Yu, Haoji Hu, Guangqun Zhang, Junguo Hu, Tao He

    With rapid urbanization, effective waste classification is a critical challenge. Traditional manual methods are time-consuming, labor-intensive, costly, and error-prone, resulting in reduced accuracy. Deep learning has revolutionized this field. Convolutional neural networks such as VGG and ResNet have dramatically improved automated sorting efficiency, and Transformer architectures like the Swin Transformer have further enhanced performance and adaptability in complex sorting scenarios. However, these approaches still struggle in complex environments and with diverse waste types, often suffering from limited recognition accuracy, poor generalization, or prohibitive computational demands. To overcome these challenges, we propose an efficient hybrid-modal fusion method, the Hybrid-modal Fusion Waste Classification Network (HFWC-Net), for precise waste image classification. HFWC-Net leverages a Transformer-based hierarchical architecture that integrates CNNs and Transformers, enhancing feature capture and fusion across varied image types for superior scalability and flexibility. By incorporating advanced techniques such as the Agent Attention mechanism and the LionBatch optimization strategy, HFWC-Net not only improves classification accuracy but also significantly reduces classification time. Comparative experimental results on the public datasets Garbage Classification, TrashNet, and our self-built MixTrash dataset demonstrate that HFWC-Net achieves Top-1 accuracy rates of 98.89%, 96.88%, and 94.35%, respectively. These findings indicate that HFWC-Net attains the highest accuracy among current methods, offering significant advantages in accelerating classification efficiency and supporting automated waste management applications.


  • Unified Stability and Plasticity for Lifelong Person Re-Identification in Cloth-Changing and Cloth-Consistent Scenarios

    by Yuming Yan, Huimin Yu, Yubin Wang, Shuyi Song, Weihu Huang, Juncan Jin

    Lifelong person re-identification (LReID) is developed for dynamic domains where domain distribution is constantly changing due to climate changes, scene changes, etc., and the data can only be collected for a specific scenario over a period of time. With the development of ReID, the issue of clothing changes has also attracted attention. Clothing change itself should be solved more from the perspective of lifelong learning because pedestrians may wear new clothes and the time span of their appearance can be long which can also cause domain changes. Meanwhile, it is difficult to know in advance whether a pedestrian is cloth-changing or cloth-consistent. However, current LReID tasks overlook these issues. To overcome these limitations, we introduce a more practical LReID task, denoted as L4C-ReID (Lifelong Person Re-Identification in Cloth-Changing and Cloth-Consistent Scenarios). This novel task empowers ReID models capable of adapting to incrementally encountered cloth-changing and cloth-consistent domains without prior knowledge of the scenario type and generalizing to unseen domains. A key challenge supposed to be fixed for LReID is the stability-plasticity dilemma. Unlike current LReID methods, which implement plasticity and stability by two contradictory loss items to achieve a sub-optimal balance, we propose an effective scheme termed Unified Stability and Plasticity (USP) that unifies these seemingly disparate concepts to achieve both harmoniously. Taking inspiration from the cognitive processes in the human brain, we decompose the cognitive processes into two independent processes: knowledge representation and knowledge operation. We then design a Knowledge Representation and Operation (KRO) framework to represent and operate the knowledge like the human brain which can better learn new knowledge and consolidate old knowledge to coordinate plasticity and stability. Additionally, we introduce Plasticizing with Stability (PWS) to generalize and optimize the learned knowledge, which integrates the implementation of plasticity and stability into one common objective item to achieve both simultaneously. To simulate the L4C-ReID setup, we gather existing cloth-changing and cloth-consistent datasets to provide a new benchmark. Extensive experiments conducted both on this new benchmark and previous benchmarks established for previous LReID setup, demonstrate the superiority of our method.


  • Open Access

    Exploring Shape Embedding for Cloth-Changing Person Re-Identification via 2D-3D Correspondences

    by Yubin Wang, Huimin Yu, Yuming Yan, Shuyi Song, Biyang Liu, Yichong Lu


  • Open Access

    A Multi-Scale Convolution and Multi-Layer Fusion Network for Remote Sensing Forest Tree Species Recognition

    by Jinjing Hou, Houkui Zhou, Junguo Hu, Huimin Yu, Haoji Hu

    Forest tree species identification in the field of remote sensing has become an important research topic. Currently, few research methods combine global and local features, making it challenging to accurately handle the similarity between different categories. Moreover, using a single deep layer for feature extraction overlooks the unique feature information at intermediate levels. This paper proposes a remote sensing image forest tree species classification method based on the Multi-Scale Convolution and Multi-Level Fusion Network (MCMFN) architecture. In the MCMFN network, the Shallow Multi-Scale Convolution Attention Combination (SMCAC) module replaces the original 7 × 7 convolution at the first layer of ResNet-50. This module uses multi-scale convolution to capture different receptive fields, and combines it with the attention mechanism to effectively enhance the ability of shallow features and obtain richer feature information. Additionally, to make efficient use of intermediate and deep-level feature information, the Multi-layer Selection Feature Fusion (MSFF) module is employed to improve classification accuracy. Experimental results on the Aerial forest dataset demonstrate a classification accuracy of 91.03%. The comprehensive experiments indicate the feasibility and effectiveness of the proposed MCMFN network.


  • Causal Intervention for Few-Shot Hypothesis Adaptation

    by Guodong Qi, Yangqi Long, Zhaohui Lu, Huimin Yu

    A domain adaptation scenario called few-shot hypothesis adaptation (FHA) only permits access to the well-trained source model and few labeled target data. Previous works may be misled by the discriminative yet undesirable domain-specific information since they learn the spurious correlation between inputs and predictions. In this work, our goal is to achieve the robust adaptation from the causality perspective. Specifically, we abstract FHA into a Structural Causal Model (SCM) and notice that the domain-specific variables are confounders, which lead to the spurious correlation. To eliminate the negative effects, we propose a novel Causal Interventional Few-shot hypothesis Adaption (CIFA) that consists of sampling, mixing and classification modules to perform backdoor adjustment. We also calculate the optimal solution for better convergence. Experiments on the benchmarks demonstrate the effectiveness of our algorithm.


  • ADfM-Net: An Adversarial Depth-From-Motion Network Based on Cross Attention and Motion Enhanced

    by Yangqi Long, Huimin Yu, Chenfeng Xu, Zhiqiang Deng

    The temporal consistent and accurate depth estimation for consecutive images is essential for many downstream applications. However, most existing methods only infer depth from a single image, ignoring the temporal information and important depth cues from motion in the sequence. Additionally, the depths of adjacent frames are estimated separately without any constraint. In this paper, we promote the temporal consistency and accuracy of depth results from the aforementioned two aspects: multi-frame framework and consistency constraint. Firstly, a framework with cross-frame attention and motion enhancement module is proposed for better temporal consistency and depth precision. Secondly, an adversarial metric learning strategy is introduced to further constrain the consistency of adjacent depth results, without any additional computation and memory cost. The experiments on KITTI and Cityscapes datasets demonstrate the effectiveness of our framework. Furthermore, noting that the traditional metrics can not reveal the consistency of the depth results, a new temporal consistency metric is proposed, which would facilitate further research.


  • GAAF: Searching Activation Functions for Binary Neural Networks Through Genetic Algorithm

    by Yanfei Li, Tong Geng, Samuel Stein, Ang Li, Huimin Yu

    Binary neural networks (BNNs) show promising utilization in cost and power-restricted domains such as edge devices and mobile systems. This is due to its significantly less computation and storage demand, but at the cost of degraded performance. To close the accuracy gap, in this paper we propose to add a complementary activation function (AF) ahead of the sign based binarization, and rely on the genetic algorithm (GA) to automatically search for the ideal AFs. These AFs can help extract extra information from the input data in the forward pass, while allowing improved gradient approximation in the backward pass. Fifteen novel AFs are identified through our GA-based search, while most of them show improved performance (up to 2.54 on ImageNet) when testing on different datasets and network models. Interestingly, periodic functions are identified as a key component for most of the discovered AFs, which rarely exist in human designed AFs. Our method offers a novel approach for designing general and application-specific BNN architecture. GAAF will be released on GitHub.