Yanan Liu

Yanan Liu

Auto-generated profile, based on publicly available information

Yanan Liu

Yunnan University, China

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
19 December 2023 in Complex & Intelligent Systems
Verified
Journal Article
Open access
PDF file
HZHao Zhang
Hao Zhang
YLYanan Liu
Yanan Liu
ZXZhaoyu Xiong
Zhaoyu Xiong
ZWZhichao Wu
Zhichao Wu
DXDan Xu
Dan Xu
Visual sentiment analysis is in great demand as it provides a computational method to recognize sentiment information in abundant visual contents from social media sites. Most of existing methods use CNNs to extract varying visual attributes for image sentiment prediction, but they failed to comprehensively consider the correlation among visual components, and are limited by the receptive field of convolutional layers as a result. In this work, we propose a visual semantic correlation network VSCNet, a Transformer-based visual sentiment prediction model. Precisely, global visual features are captured through an extended attention network stacked by a well-designed extended attention mechanism like Transformer. An off-the-shelf object query tool is used to determine the local candidates of potential affective regions, by which redundant and noisy visual proposals are filtered out. All candidates considered affective are embedded into a computable semantic space. Finally, a fusion strategy integrates semantic representations and visual features for sentiment analysis. Extensive experiments reveal that our method outperforms previous studies on 5 annotated public image sentiment datasets without any training tricks. More specifically, it achieves 1.8% higher accuracy on FI benchmark compared with other state-of-the-art methods.

8 May 2024 in IEEE Transactions on Circuits and Systems for Video Technology
Verified
Journal Article
Closed access
YLYanan Liu
Yanan Liu
YLYanqiu Li
Yanqiu Li
HZHao Zhang
Hao Zhang
XZXuejie Zhang
Xuejie Zhang
DXDan Xu
Dan Xu
Skeleton-based action recognition has broad prospects owing to the fact that skeleton data is more robust to scene noise and camera view changes. Recently, researchers mainly aim to explore deep-learning feature engineering with competitive recognition accuracy for skeleton actions. However, a high-performance recognition network is usually stacked by complex feature extraction modules introducing massive computational costs. In this work, we designed a powerful and universal action knowledge distillation paradigm based on decoupled knowledge distillation for transferring action knowledge from heavy teachers to lightweight students more robustly. We constructed a network architecture space consisting of the shrinking versions of outdated 2s-AGCN and searched for several robust students. On this basis, this paradigm is further developed into a powerful decoupled knowledge embedded graph convolutional network (DKE-GCN), which outperforms the teacher significantly on three public datasets and achieves the state-of-the-art. In addition, a light-DKE-GCN is designed to achieve comparable performance with teacher with 16× less parameters, 26× less FLOPs and 8× FPS.

24 May 2024
Verified
Conference Paper
Open access
SSShouzheng Sun
Shouzheng Sun
QTQianhan Tang
Qianhan Tang
YLYanan Liu
Yanan Liu
HZHao Zhang
Hao Zhang
QSQianyun Song
Qianyun Song
DXDan Xu
Dan Xu
This paper propose a novel dance action dataset – YNU-Dance. To preserve and inherit ethnic dances, we have collected and constructed a dataset of ethnic dance actions. The dataset encompasses unique dances from 10 different ethnic groups, including multi-view synchronized RGB images and precise motion data captured using motion capture devices. The dataset comprises 256320 frames of RGB images, with a total video duration of 1 hour and 11 minutes, and consists of 3402 action data segments. We compare the YNU-Dance dataset with other publicly available 3D action recognition datasets, highlighting its uniqueness and importance in the field of dance action. Furthermore, we evaluate the YNU-Dance dataset using skeleton action recognition methods and introduces a Prototypical Spatial Temporal model based on prototype networks (ProtoST) to address the small-sample challenge and achieve improved ethnic dance action classification. Dance action data is segmented into time segments suitable for the current model, and data augmentation is applied to enhance model generalization performance. The experiments show that the YNU-Dance dataset, with its multimodal and multi-view characteristics, provides a more complex challenge for action recognition models. This emphasizes the practical research value of the ethnic dance dataset we have constructed and encourages further in-depth studies on this dataset. The video and motion data will be made available for download in the subsequent stages.

6 January 2023
Verified
Conference Paper
Open access
XHXiao Han
Xiao Han
HZHao Zhang
Hao Zhang
ZXZhaoyu Xiong
Zhaoyu Xiong
YLYanan Liu
Yanan Liu
KHKangjian He
Kangjian He
DXDan Xu
Dan Xu
Image emotion analysis aims to determine an author's attitude toward the overall contextual polarity of a topic or document. It is an instrumental but challenging task, especially when analyzing fine-grained emotions. Existing techniques based on convolutional neural networks learn affective representations from local visual features while ignoring that both the global semantics of images and objects impact emotional representation learning. To address this limitation, we propose a unique Coupling Dependent Attentive Model (CDA) for image emotion analysis. The model extracts image semantics through several residual blocks and transforms the semantic feature maps obtained from fully convolutional networks through spatial attention. It then implants a self-attention mechanism to mine the correlations between local visual features and achieve the final affective representation. Finally, considering the affective image, we proposed an automated image augmentation proxy to increase model generalization. Extensive experiments on six benchmark datasets show that the proposed model outperforms state-of-the-art methods in image emotion analysis.