ZLZun Li

Zun Li

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Zun Li has a total of 42 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Consensus-Driven Coarse-to-Fine Consistency Learning for Co-Salient Object Detection

    by Zun Li, Jiashuai Xie, Shaoyue Song, Jian Zhao, Lifang Wu

    Co-salient object detection aims to identify common salient objects within related group images. The key challenge involves effectively capturing and utilizing the intrinsic consensus across group representations. Existing methods usually mine consensus from global high-level semantics and inject it into basic multi-level features, yet suffer from two key challenges: incomplete consensus exploration due to limited structural modeling under scale/position variations, and non-consensus noise introduced during consensus utilization. In this paper, we propose a novel consensus-driven coarse-to-fine consistency learning method named CCFC, which explores and exploits semantic consensus to achieve multi-level consistent salient features in a coarse-to-fine learning manner. Specifically, we first design a semantic consensus mining module that jointly interacts local structures and global semantics to capture semantic consensus. Then, under the guidance of such consensus, we develop a consensus-driven feature interaction module, which discovers fine-grained low-level and mid-level consistent salient features with cross-attention mechanism. Collaborating with these two modules, the proposed CCFC adaptively propagates semantic consensus across each level consistent features, preserving discriminative consistent salient cues while filtering out redundant noise, and thus predicts co-salient objects with complete contents and clear boundaries. Extensive experiments on three challenging benchmark datasets verify that CCFC achieves superior performance over state-of-the-art methods, with 0.797 \(S_{\alpha}\) and 0.065 MAE on the challenging CoCA dataset. Source code is available at: https://github.com/xiye2002/CCFC.

  • Open Access

    VicKAM: Visual Conceptual Knowledge Guided Action Map for Weakly Supervised Group Activity Recognition

    by Zhuming Wang, Yihao Zheng, Jiarui Li, Yaofei Wu, Yan Huang, Zun Li, Lifang Wu, Liang Wang


  • Quality-Invariant Domain Generalization for Face Anti-Spoofing

    by Yongluo Liu, Zun Li, Yaowen Xu, Zhizhi Guo, Zhaofan Zou, Lifang Wu


  • Open Access

    Knowledge Guided Entity-aware Video Captioning and A Basketball Benchmark

    by Zeyu Xi, Ge Shi, Xuefen Li, Junchi Yan, Zun Li, Lifang Wu, Zilin Liu, Liang Wang

    Despite the recent emergence of video captioning models, how to generate the text description with specific entity names and fine-grained actions is far from being solved, which however has great applications such as basketball live text broadcast. In this paper, a new multimodal knowledge graph supported basketball benchmark for video captioning is proposed. Specifically, we construct a multimodal basketball game knowledge graph (KG_NBA_2022) to provide additional knowledge beyond videos. Then, a multimodal basketball game video captioning (VC_NBA_2022) dataset that contains 9 types of fine-grained shooting events and 286 players' knowledge (i.e., images and names) is constructed based on KG_NBA_2022. We develop a knowledge guided entity-aware video captioning network (KEANet) based on a candidate player list in encoder-decoder form for basketball live text broadcast. The temporal contextual information in video is encoded by introducing the bi-directional GRU (Bi-GRU) module. And the entity-aware module is designed to model the relationships among the players and highlight the key players. Extensive experiments on multiple sports benchmarks demonstrate that KEANet effectively leverages extera knowledge and outperforms advanced video captioning models. The proposed dataset and corresponding codes will be publicly available soon


  • Open Access

    Multi-Stage Manufacturing for Preoperative Medical Models with Overhanging Components

    by Mingli Xiang, Zun Li, Lin Lu, Lifang Wu

    Medical models play an instrumental role in replicating a patient’s unique anatomy, enhancing surgical outcomes, and fostering effective communication between doctors and patients. While 3D printing technology offers bespoke solutions for such models, producing designs with multiple overhanging tissues often requires advanced 3D printers – making it unattainable for conventional FDM fabrication methods. To bridge this gap, we present a cost-efficient, multi-stage, printing-molding hybrid manufacturing technique that incrementally solidifies complex medical models. Leveraging our adaptive optimization algorithm, we determine the optimal molding direction and the most streamlined manufacturing process with the fewest stages. As a testament to our method’s efficacy, we successfully printed a liver model featuring overhanging tumors.

  • Open Access

    Improving Face Anti-spoofing via Advanced Multi-perspective Feature Learning

    by Zhuming Wang, Yaowen Xu, Lifang Wu, Hu Han, Yukun Ma, Zun Li

    Face anti-spoofing (FAS) plays a vital role in securing face recognition systems. Previous approaches usually learn spoofing features from a single perspective, in which only universal cues shared by all attack types are explored. However, such single-perspective-based approaches ignore the differences among various attacks and commonness between certain attacks and bona fides, thus tending to neglect some non-universal cues that contain strong discernibility against certain types. As a result, when dealing with multiple types of attacks, the above approaches may suffer from the uncomprehensive representation of bona fides and spoof faces. In this work, we propose a novel Advanced Multi-Perspective Feature Learning network (AMPFL), in which multiple perspectives are adopted to learn discriminative features, to improve the performance of FAS. Specifically, the proposed network first learns universal cues and several perspective-specific cues from multiple perspectives, then aggregates the above features and further enhances them to perform face anti-spoofing. In this way, AMPFL obtains features that are difficult to be captured by single-perspective-based methods and provides more comprehensive information on bona fides and spoof faces, thus achieving better performance for FAS. Experimental results show that our AMPFL achieves promising results in public databases, and it effectively solves the issues of single-perspective-based approaches.


  • Open Access

    A novel scheme to developing printing protocol with the required surface roughness

    by Zening Men, Lidong Zhao, Feng Yang, Mengcheng Jiang, Lifang Wu, Zun Li

    3D printing has the characteristics of layer-by-layer manufacturing, which causes high surface roughness. This paper proposes a scheme to develop a printing protocol combining layer-wise and continuous printing based on the required surface roughness. It involves four modules: model slicing, printing pattern estimation, slices adjusting and printing protocol generation. Firstly, the candidate slices are obtained by model slicing based on the required surface roughness. Secondly, the printing pattern of each slice is estimated based on the max-min distance of the corresponding slice and Maximum Filled Distance (MFD) of the printing resin material. Then, the candidate slices are adjusted based on the maximum and minimum printable thickness of the printer. Finally, the printing protocol involving slice number, slice thickness, printing pattern and printing time is generated, and the surface roughness of the printed objects using the generated printing protocol can be estimated. Two printing protocols of the model cup with different required surface roughness are automatically generated. And two objects are printed based on the corresponding printing protocols. The roughness of the printed objects is measured using the roughness tester. The average roughness of the printed objects is smaller than the required roughness because the roughness of continuous printing is small. And the error between the measured and the predicted roughness is smaller than 2 µm.

  • Deep spatio-frequency saliency detection

    by Zun Li, Congyan Lang, Tao Wang, Yidong Li, Jiashi Feng


  • Dense Attentive Feature Enhancement for Salient Object Detection

    by Zun Li, Congyan Lang, Liqian Liang, Jian Zhao, Songhe Feng, Qibin Hou, Jiashi Feng

    Attention mechanisms have been proven highly effective for salient object detection. Most previous works utilize attention as a self-gated module to reweigh the feature maps at different levels independently. However, they are limited to certain-level guidance and could not satisfy the need of both accurately detecting intact objects and maintaining their detailed boundaries. In this paper, we build dense attention upon features from multiple levels simultaneously and propose a novel Dense Attentive Feature Enhancement (DAFE) module for efficient feature enhancement in saliency detection. DAFE stacks several attentional units and densely connects attentive feature output from current unit to its all subsequent units. This allows feature maps at deep units to absorb attentive information from shallow units, thus more discriminative information can be efficiently selected at the final output. Note that DAFE is plug and play, which can be effortlessly inserted into any saliency or video saliency models for their performance improvements. We further instantiate a highly effective Dense Attentive Feature Enhancement Network (DAFE-Net) for accurate salient object detection. DAFE-Net constructs DAFE over the aggregation feature that contains both semantics and saliency details, the entire salient objects and their boundaries can be well retained through dense attentions. Extensive experiments demonstrate that the proposed DAFE module is highly effective, and the DAFE-Net performs favorably compared with state-of-the-art approaches.


  • Cross-Layer Feature Pyramid Network for Salient Object Detection

    by Zun Li, Congyan Lang, Jun Hao Liew, Yidong Li, Qibin Hou, Jiashi Feng

    Feature pyramid network (FPN) based models, which fuse the semantics and salient details in a progressive manner, have been proven highly effective in salient object detection. However, it is observed that these models often generate saliency maps with incomplete object structures or unclear object boundaries, due to the indirect information propagation among distant layers that makes such fusion structure less effective. In this work, we propose a novel Cross-layer Feature Pyramid Network (CFPN), in which direct cross-layer communication is enabled to improve the progressive fusion in salient object detection. Specifically, the proposed network first aggregates multi-scale features from different layers into feature maps that have access to both the high- and low- level information. Then, it distributes the aggregated features to all the involved layers to gain access to richer context. In this way, the distributed features per layer own both semantics and salient details from all other layers simultaneously, and suffer reduced loss of important information during the progressive feature fusion. At last, CFPN fuses the distributed features of each layer stage-by-stage. This way, the high-level features that contain context useful for locating complete objects are preserved until the final output layer, and the low-level features that contain spatial structure details are embedded into each layer to preserve spatial structural details. Extensive experimental results over six widely used salient object detection benchmarks and with three popular backbones clearly demonstrate that CFPN can accurately locate fairly complete salient regions and effectively segment the object boundaries.