Author name cluster

Nannan Li

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

8 papers

1 author row

EAAI Journal 2026 Journal Article

Global frequency-aware multi-scale feature learning for point cloud normal estimation

Wei Jin
Jun Zhou
Nannan Li
Xiuping Liu

Estimating accurate surface normals from point clouds remains a core challenge in three-dimensional (3D) computer vision due to irregular sampling and the difficulty of modeling global geometric context. In this paper, we propose a frequency-domain learning framework that addresses these issues by preserving global information throughout the feature extraction process. Specifically, we introduce a Fourier-Based Multi-Branch Patch Refinement Module at the data level to enhance patch representation with spectral cues, and a Fourier-Based Feature Refinement Layer to integrate local and global geometric features. A multi-scale fusion strategy is further adopted to ensure hierarchical consistency across resolutions. Compared to existing spatial-domain strategies, our method improves global context awareness by incorporating frequency-domain information, effectively mitigating the loss of global features commonly introduced during early-stage local convolutional operations. Experimental results demonstrate consistent performance improvements over prior methods, with gains of 1. 0% on the Point Cloud Property Network (PCPNet) dataset, which is a benchmark for learning local 3D shape properties from raw point clouds, 0. 76% on the Famous Shape dataset (FamousShape), which consists of several well-known 3D mesh models such as the Utah Teapot and Stanford Bunny, and 0. 65% on the Scene Meshes dataset with Annotations (SceneNN), which is a richly annotated collection of indoor 3D scenes.

Details DOI

JBHI Journal 2026 Journal Article

USCNet: Transformer-Based Multimodal Fusion with Segmentation Guidance for Urolithiasis Classification

Changmiao Wang
Songqi Zhang
Yongquan Zhang
Yifei Wang
Liya Liu
Nannan Li
Xingzhi Li
Jiexin Pan

Kidney stone disease ranks among the most prevalent conditions in urology, and understanding the composition of these stones is essential for creating personalized treatment plans and preventing recurrence. Current methods for analyzing kidney stones depend on post operative specimens, which prevents rapid classification before surgery. To overcome this limitation, we introduce a new approach called the Urinary Stone Segmentation and Classification Network (USCNet). This innovative method allows for precise preoperative classification of kidney stones by integrating Computed Tomography (CT) images with clinical data from Electronic Health Records (EHR). USCNet employs a Transformer-based multimodal fusion framework with CT-EHR attention and segmentation-guided attention modules for accurate classification. Moreover, a dynamic loss function is introduced to effectively balance the dual objectives of segmentation and classification. Experiments on an in-house kidney stone dataset show that USCNet demonstrates outstanding performance across all evaluation metrics, with its classification efficacy significantly surpassing existing mainstream methods. This study presents a promising solution for the precise preoperative classification of kidney stones, offering substantial clinical benefits. The source code has been made publicly available: https://github.com/fancccc/KidneyStoneSC.

Details DOI

EAAI Journal 2025 Journal Article

Causality thinking for large-scale long-tailed video action recognition

Zhengjin Zhang
Nannan Li
Wenmin Wang
Huiwen Guo
Wei Jin
Sudan Huang

Video action recognition aims to accurately classify actions in video data by leveraging spatial–temporal representations. However, this task faces two major challenges: (1) the intrinsic long-tailed data distribution in real-world scenarios, which skews model learning toward overrepresented classes, and (2) the presence of spurious correlations in the data, which undermines prediction reliability. In this paper, we propose a causal inference framework for long-tailed video action recognition. A novel debiasing approach with Causal Intervention and Counterfactual Reasoning (CICR) is proposed to yield more robust predictions. The Structural Causal Model (SCM) is constructed to identify and analyze the causal relationships among spatial–temporal variables and long-tailed class distributions. Specifically, we first diagnose momentum as the confounder and use the intervention to eliminate the spurious relationship existed among video frame appearance features. The counterfactual reasoning is then utilized during inference to calculate the direct causal effect more accurately. During experiments conducted on two large-scale datasets, the proposed approach improved video action recognition accuracy by 1. 9% on EPIC-KITCHENS-100 and 3. 65% on Something-Something-V2-LT, thereby showing the effectiveness of the CICR model in mitigating spurious correlation and long-tailed distribution issues. Codes are available at https: //github. com/sandyzhang2021/Causality-for-long-tailed-video-action-recognition.

Details DOI

JBHI Journal 2024 Journal Article

Skin Conductance-Based Acupoint and Non-Acupoint Recognition Using Machine Learning

Feifei Shi
Huansheng Ning
Ruoxiu Xiao
Tao Zhu
Nannan Li

Acupoints (APs) prove to have positive effects on disease diagnosis and treatment, while intelligent techniques for the automatic detection of APs are not yet mature, making them more dependent on manual positioning. In this paper, we realize the skin conductance-based APs and non-APs recognition with machine learning, which could assist in APs detection and localization in clinical practice. Firstly, we collect skin conductance of traditional Five-Shu Point and their corresponding non-APs with wearable sensors, establishing a dataset containing over 36000 samples of 12 different AP types. Then, electrical features are extracted from the time domain, frequency domain, and nonlinear perspective respectively, following which typical machine learning algorithms (SVM, RF, KNN, NB, and XGBoost) are demonstrated to recognize APs and non-APs. The results demonstrate XGBoost with the best precision of 66. 38%. Moreover, we also quantify the impacts of the differences among AP types and individuals, and propose a pairwise feature generation method to weaken the impacts on recognition precision. By using generated pairwise features, the recognition precision could be improved by 7. 17%. The research systematically realizes the automatic recognition of APs and non-APs, and is conducive to pushing forward the intelligent development of APs and Traditional Chinese Medicine theories.

Details DOI

EAAI Journal 2024 Journal Article

Text-guided Graph Temporal Modeling for few-shot video classification

Fuqin Deng
Jiaming Zhong
Nannan Li
Lanhui Fu
Bingchun Jiang
Yi Ningbo
Feng Qi
He Xin

Large-scale pre-trained models and graph neural networks have recently demonstrated remarkable success in few-shot video classification tasks. However, they generally suffer from two key limitations: i) the temporal relations between adjacent frames tends to be ambiguous due to the lack of explicit temporal modeling. ii) the absence of multi-modal semantic knowledge in query videos results in inaccurate prototypes construction and an inability to achieve multi-modal temporal alignment metrics. To address these issues, we develop a Text-guided Graph Temporal Modeling (TgGTM) method that consists of two crucial components: a text-guided feature refinement module and a learnable Query text-token contrastive objective. Specifically, the former leverages the Temporal masking layer to guide the model in learning temporal relationships between adjacent frames. Additionally, it utilizes multi-modal information to refine video prototypes for comprehensive few-shot video classification. The latter addresses the feature discrepancy between multi-modal support features and single-modal query features by aligning a learnable Query text-token with corresponding base class text descriptions. Extensive experiments on four commonly used benchmarks demonstrate the effectiveness of our proposed method, which achieves mean accuracies of 54. 4%, 80. 3%, 91. 9%, and 96. 2% for 5-way 1-shot classification on SSV2-Small, HMDB51, Kinetics, and UCF101, respectively. These results are superior compared to existing state-of-the-art methods. A detailed ablation showcases the importance of learning temporal relationships between adjacent frames and obtaining Query text-token. The source code and models will be publicly available at https: //github. com/JiaMingZhong2621/TgGTM.

Details DOI

AAAI Conference 2019 Conference Paper

Meta Learning for Image Captioning

Nannan Li
Zhenzhong Chen
Shan Liu

Reinforcement learning (RL) has shown its advantages in image captioning by optimizing the non-differentiable metric directly in the reward learning process. However, due to the reward hacking problem in RL, maximizing reward may not lead to better quality of the caption, especially from the aspects of propositional content and distinctiveness. In this work, we propose to use a new learning method, meta learning, to utilize supervision from the ground truth whilst optimizing the reward function in RL. To improve the propositional content and the distinctiveness of the generated captions, the proposed model provides the global optimal solution by taking different gradient steps towards the supervision task and the reinforcement task, simultaneously. Experimental results on MS COCO validate the effectiveness of our approach when compared with the state-of-the-art methods.

PDF Details

IJCAI Conference 2018 Conference Paper

Image Cationing with Visual-Semantic LSTM

Nannan Li
Zhenzhong Chen

In this paper, a novel image captioning approach is proposed to describe the content of images. Inspired by the visual processing of our cognitive system, we propose a visual-semantic LSTM model to locate the attention objects with their low-level features in the visual cell, and then successively extract high-level semantic features in the semantic cell. In addition, a state perturbation term is introduced to the word sampling strategy in the REINFORCE based method to explore proper vocabularies in the training process. Experimental results on MS COCO and Flickr30K validate the effectiveness of our approach when compared to the state-of-the-art methods.

PDF Details

AAAI Conference 2018 Conference Paper

SAP: Self-Adaptive Proposal Model for Temporal Action Detection Based on Reinforcement Learning

Jingjia Huang
Nannan Li
Tao Zhang
Ge Li
Tiejun Huang
Wen Gao

Existing action detection algorithms usually generate action proposals through an extensive search over the video at multiple temporal scales, which brings about huge computational overhead and deviates from the human perception procedure. We argue that the process of detecting actions should be naturally one of observation and reﬁnement: observe the current window and reﬁne the span of attended window to cover true action regions. In this paper, we propose a Self-Adaptive Proposal (SAP) model that learns to ﬁnd actions through continuously adjusting the temporal bounds in a self-adaptive way. The whole process can be deemed as an agent, which is ﬁrstly placed at the beginning of the video and traverse the whole video by adopting a sequence of transformations on the current attended region to discover actions according to a learned policy. We utilize reinforcement learning, especially the Deep Q-learning algorithm to learn the agent’s decision policy. In addition, we use temporal pooling operation to extract more effective feature representation for the long temporal window, and design a regression network to adjust the position offsets between predicted results and the ground truth. Experiment results on THUMOS’14 validate the effectiveness of SAP, which can achieve competitive performance with current action detection algorithms via much fewer proposals.

PDF Details