ZZZeyang Zhao

Zeyang Zhao

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Zeyang Zhao has a total of 22 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Empowering Agrifood System with Artificial Intelligence: A Survey of the Progress, Challenges and Opportunities

    by Tao Chen, Liang Lv, Di Wang, Jing Zhang, Yue Yang, Zeyang Zhao, Chen Wang, Xiaowei Guo, Hao Chen, Qingye Wang, et al.

    With the world population rapidly increasing, transforming our agrifood systems to be more productive, efficient, safe, and sustainable is crucial to mitigate potential food shortages. Recently, artificial intelligence (AI) techniques such as deep learning (DL) have demonstrated their strong abilities in various areas, including language, vision, remote sensing (RS), and agrifood systems applications. However, the overall impact of AI on agrifood systems remains unclear. In this article, we thoroughly review how AI techniques can transform agrifood systems and contribute to the modern agrifood industry. First, we summarize the data acquisition methods in agrifood systems, including acquisition, storage, and processing techniques. Second, we present a progress review of AI methods in agrifood systems, specifically in agriculture, animal husbandry, and fishery, covering topics such as agrifood classification, growth monitoring, yield prediction, and quality assessment. Furthermore, we highlight potential challenges and promising research opportunities for transforming modern agrifood systems with AI. We hope this survey can offer an overall picture to newcomers in the field and serve as a starting point for their further research. The project website is https://github.com/Frenkie14/Agrifood-Survey.


  • Top Cited Papers

    ARTrackV2: Prompting Autoregressive Tracker Where to Look and How to Describe

    by Yifan Bai, Zeyang Zhao, Yihong Gong, Xing Wei

    We present ARTrackV2, which integrates two pivotal aspects of tracking: determining where to look (localization) and how to describe (appearance analysis) the target object across video frames. Building on the foundation of its predecessor, ARTrackV2 extends the concept by introducing a unified generative framework to “read out” object's trajectory and “retell” its appearance in an autoregressive manner. This approach fosters a time-continuous methodology that models the joint evolution of motion and visual features, guided by previous estimates. Furthermore, ARTrackV2 stands out for its efficiency and simplicity, obviating the less efficient intra-frame autoregression and hand-tuned parameters for appearance updates. Despite its simplicity, ARTrackV2 achieves state-of-the-art performance on prevailing benchmark datasets while demonstrating a remarkable efficiency improvement. In particular, ARTrackV2 achieves an AO score of 79. 5% on GOT-10k and an AUC of 86. 1% on TrackingNet while being 3.6× faster than ARTrack.


  • Top Cited Papers
    Open Access

    Landslide Susceptibility Mapping Considering Landslide Local-Global Features Based on CNN and Transformer

    by Zeyang Zhao, Tao Chen, Jie Dou, Gang Liu, Antonio Plaza

    Landslide susceptibility mapping (LSM) is a crucial step in quantitatively assessing landslide risk, essential for geologic hazards prevention. With the rapid development of deep learning models, convolutional neural networks (CNN) and Transformer architectures have been applied to LSM. However, these models still face the challenges of suboptimal mapping accuracy and limited capacity for multi-level landslide features extraction. In this study, we present a CNN-Transformer Local-Global Feature Extraction Network (CTLGNet) that combines the strengths of both CNN and Transformer models to effectively extract both landslide local and global features. We apply this model to LSM in two regions: the Three Gorges Reservoir area and Jiuzhaigou. To begin, nine landslide conditioning factors are selected and analyzed to construct the landslide dataset for LSM. Subsequently, the dataset is randomly split into training, validation, and test datasets in a 6:2:2 ratio to attain LSM results. Then, CTLGNet is compared to CNN, residual neural network (ResNet), densely connected convolutional network (DenseNet), Vision Transformer (ViT), and Fractional Fourier Image Transformer (FrIT) using various evaluation metrics. The results demonstrate that CTLGNet exhibits exceptional landslide prediction and generalization capabilities, outperforming the other five models across all evaluation metrics except Recall, with AUC values of 0.9817 and 0.9693 for the two regions respectively. The LSM results indicate that CTLGNet can effectively extract both landslide local and global features to achieve landslide localization and detail capture. Overall, our proposed framework excels in extracting multi-level landslide features and holds great potential for widespread application.


  • Open Access

    LCFSTE: Landslide Conditioning Factors and Swin Transformer Ensemble for Landslide Susceptibility Assessment

    by Tao Chen, Qingye Wang, Zeyang Zhao, Gang Liu, Jie Dou, Antonio Plaza

    Landslide susceptibility assessment (LSA) holds crucial importance in guiding regional disaster prevention and reduction efforts. However, current deep learning (DL) models for LSA encounter challenges like insufficient landslide data samples and uneven distribution. In this paper, we develop a new hybrid framework named LCFSTE, which integrates landslide conditioning factors (LCFs) and Swin Transformer (Swin-T) for LSA. With this framework, we fully leverage the powerful nonlinear feature extraction capability of Swin-T to extract abstract features from both landslides and LCFs. This approach ultimately enhances the precision and reliability of LSA. To assess the performance of our newly proposed framework, we selected Jiuzhaigou County, China, as our study area. Firstly, a dataset for LSA was constructed using historical landslide data and 11 multi-source LCFs. Then, these factors were screened through multicollinearity test and factor importance analysis using variance inflation factors, tolerance, and information gain rate. Subsequently, the dataset was divided into three subsets: 60% for training, 20% for validation and 20% for testing. Then, the LSA results were compared with four DL models. Seven evaluation metrics (EMs) are chosen to quantitatively evaluate the performance of these five LSA models. The results demonstrated that, among these seven EMs, LCFSTE outperformed the others, achieving the highest score in six out of the seven considered EMs. This outcome highlights the promising applicability of LCFSTE in enhancing LSA accuracy.


  • Top Cited Papers
    Open Access

    BisDeNet: A New Lightweight Deep Learning-Based Framework for Efficient Landslide Detection

    by Tao Chen, Xiao Gao, Gang Liu, Chen Wang, Zeyang Zhao, Jie Dou, Ruiqing Niu, Antonio J. Plaza

    Landslides are catastrophic geological events that can cause significant damage to properties and result in the loss of human lives. Deep-learning technology applied to optical remote sensing images can enable effective landslide-prone area detection. However, conventional landslide detection (LD) models often employ complex structural designs to ensure detection accuracy. The complexity often hampers the detection speed, rendering these models inadequate for the swift emergency monitoring of landslides. To address these problems, we propose a new lightweight deep-learning-based framework, BisDeNet, for efficient LD. To improve the efficiency of the proposed BisDeNet, we replaced the context path in the original BiSeNet with DenseNet due to its strong feature extraction ability, few required parameters, and low model complexity. Two sites with different and representative landslide developments were selected as the study areas to verify the performance of our proposed BisDeNet. Additionally, we introduced landslide causative factors to enhance the sampling dataset. To evaluate the effectiveness of our approach, we compared the performance of our BisDeNet with the performances of three other BiSeNet-based methods and an advanced transformer-based model data-efficient image transformer (DeiT). Our experimental results indicate that the F 1-scores of BisDeNet in the two study areas are 0.9006 and 0.8850, which are 26.22% and 1.86% higher than the scores of BiSeNet, respectively, but slightly lower than that of the DeiT model. Furthermore, our proposed BisDeNet requires the fewest number of parameters and the least memory out of the five models.