Xiangfeng Shen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xiangfeng Shen has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Nianfeng Li
College of Computer Science and Technology, Changchun University, Changchun 130022, China
4 shared publications Jia Tian
School of Computer Science and Technology, Changchun University, Changchun, Jilin, China
2 shared publicationsTong Guan
School of Computer Science and Technology, Changchun University, Changchun, Jilin, China
2 shared publications- Zhiguo Xiao
College of Computer Science and Technology, Changchun University, Changchun 130022, China
2 shared publications - Haiyue Zhao
College of Journalism and Communication, Changchun University of Technology, Changchun, China
1 shared publication - Jia Tian
College of Computer Science and Technology, Changchun University, Changchun 130022, China
1 shared publication
Latest Publications
- Open Access
- 2 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 6277-6291
TransDeep: Transformer-Integrated DeepLabV3+ for Image Semantic Segmentation
In recent years, image semantic segmentation algorithms have made significant progress driven by deep learning technology, and are widely used in fields such as medical image analysis, assistive technology for the blind, and autonomous driving. Aiming at problems such as the inability of many image segmentation algorithms to fully capture global context information, low computational efficiency, and insufficient context information fusion. This article integrates the Transformer mechanism and CA mechanism into the DeepLabV3+ network and proposes the TransDeep network. In the encoder, two different backbones, Xception and MobileNetV2, are first used for feature extraction, and a better backbone network is selected. Furthermore, based on the lightweight backbone network, the Transformer mechanism is integrated into the advanced features of the backbone to enhance long-range dependence. Secondly, the Coord Attention module (CA) is added after the low-level features of the backbone to strengthen information such as edge and detail features. Finally, the Coord Attention mechanism (CA) is added after the ASPP module to allow the model to focus on key image features while effectively filtering out irrelevant background information. Experimental results show that the TransDeep network can improve the accuracy of key categories and effectively improve the network’s segmentation accuracy of targets in images. It achieved an MIoU of 73.5% on the Pascal test set and a good performance of 79.95% on the CamVid test set. Compared with the baseline model, they improved by 2.10% and 2.61% respectively.
- 2 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 6277-6291
Open AccessTransDeep: Transformer-Integrated DeepLabV3+ for Image Semantic Segmentation
In recent years, image semantic segmentation algorithms have made significant progress driven by deep learning technology, and are widely used in fields such as medical image analysis, assistive technology for the blind, and autonomous driving. Aiming at problems such as the inability of many image segmentation algorithms to fully capture global context information, low computational efficiency, and insufficient context information fusion. This article integrates the Transformer mechanism and CA mechanism into the DeepLabV3+ network and proposes the TransDeep network. In the encoder, two different backbones, Xception and MobileNetV2, are first used for feature extraction, and a better backbone network is selected. Furthermore, based on the lightweight backbone network, the Transformer mechanism is integrated into the advanced features of the backbone to enhance long-range dependence. Secondly, the Coord Attention module (CA) is added after the low-level features of the backbone to strengthen information such as edge and detail features. Finally, the Coord Attention mechanism (CA) is added after the ASPP module to allow the model to focus on key image features while effectively filtering out irrelevant background information. Experimental results show that the TransDeep network can improve the accuracy of key categories and effectively improve the network’s segmentation accuracy of targets in images. It achieved an MIoU of 73.5% on the Pascal test set and a good performance of 79.95% on the CamVid test set. Compared with the baseline model, they improved by 2.10% and 2.61% respectively.
- Open Access
- 12 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 175715-175741
Research Advances in Deep Learning for Image Semantic Segmentation Techniques
Image semantic segmentation represents a significant area of research within the field of computer vision. With the advent of deep learning, image semantic segmentation techniques that integrate deep learning have demonstrated superior accuracy compared to traditional image semantic segmentation methods. Recently, the Mamba architecture has demonstrated superior semantic segmentation performance compared to the Transformer architecture, and has consequently become a research focus in this field. Nevertheless, the specifics of the Mamba architecture have remained underexplored in the extant literature. This review provides a comprehensive overview of the latest research progress in deep learning techniques for semantic segmentation. It offers a systematic review of traditional convolutional neural network (CNN)-based architectures and focuses on a series of emerging architectures, including the Transformer architecture, the Mamba architecture, and cutting-edge approaches such as self-supervised learning strategies. For each category, a detailed account is provided of the principal algorithms and techniques employed, together with a report on the performance achieved using datasets commonly used in the field.
- 12 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 175715-175741
Open AccessResearch Advances in Deep Learning for Image Semantic Segmentation Techniques
Image semantic segmentation represents a significant area of research within the field of computer vision. With the advent of deep learning, image semantic segmentation techniques that integrate deep learning have demonstrated superior accuracy compared to traditional image semantic segmentation methods. Recently, the Mamba architecture has demonstrated superior semantic segmentation performance compared to the Transformer architecture, and has consequently become a research focus in this field. Nevertheless, the specifics of the Mamba architecture have remained underexplored in the extant literature. This review provides a comprehensive overview of the latest research progress in deep learning techniques for semantic segmentation. It offers a systematic review of traditional convolutional neural network (CNN)-based architectures and focuses on a series of emerging architectures, including the Transformer architecture, the Mamba architecture, and cutting-edge approaches such as self-supervised learning strategies. For each category, a detailed account is provided of the principal algorithms and techniques employed, together with a report on the performance achieved using datasets commonly used in the field.
- Open Access
Dense Pedestrian Detection Based on GR-YOLO
In large public places such as railway stations and airports, dense pedestrian detection is important for safety and security. Deep learning methods provide relatively effective solutions but still face problems such as feature extraction difficulties, image multi-scale variations, and high leakage detection rates, which bring great challenges to the research in this field. In this paper, we propose an improved dense pedestrian detection algorithm GR-yolo based on Yolov8. GR-yolo introduces the repc3 module to optimize the backbone network, which enhances the ability of feature extraction, adopts the aggregation–distribution mechanism to reconstruct the yolov8 neck structure, fuses multi-level information, achieves a more efficient exchange of information, and enhances the detection ability of the model. Meanwhile, the Giou loss calculation is used to help GR-yolo converge better, improve the detection accuracy of the target position, and reduce missed detection. Experiments show that GR-yolo has improved detection performance over yolov8, with a 3.1% improvement in detection means accuracy on the wider people dataset, 7.2% on the crowd human dataset, and 11.7% on the people detection images dataset. Therefore, the proposed GR-yolo algorithm is suitable for dense, multi-scale, and scene-variable pedestrian detection, and the improvement also provides a new idea to solve dense pedestrian detection in real scenes.
Open AccessDense Pedestrian Detection Based on GR-YOLO
In large public places such as railway stations and airports, dense pedestrian detection is important for safety and security. Deep learning methods provide relatively effective solutions but still face problems such as feature extraction difficulties, image multi-scale variations, and high leakage detection rates, which bring great challenges to the research in this field. In this paper, we propose an improved dense pedestrian detection algorithm GR-yolo based on Yolov8. GR-yolo introduces the repc3 module to optimize the backbone network, which enhances the ability of feature extraction, adopts the aggregation–distribution mechanism to reconstruct the yolov8 neck structure, fuses multi-level information, achieves a more efficient exchange of information, and enhances the detection ability of the model. Meanwhile, the Giou loss calculation is used to help GR-yolo converge better, improve the detection accuracy of the target position, and reduce missed detection. Experiments show that GR-yolo has improved detection performance over yolov8, with a 3.1% improvement in detection means accuracy on the wider people dataset, 7.2% on the crowd human dataset, and 11.7% on the people detection images dataset. Therefore, the proposed GR-yolo algorithm is suitable for dense, multi-scale, and scene-variable pedestrian detection, and the improvement also provides a new idea to solve dense pedestrian detection in real scenes.
- Open Access
- 16 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 55225-55233
China University Online Public Opinion Risk Dataset
With the widespread popularity of social software and self-media, online public opinion incidents in colleges and universities occur frequently and present a complicated situation. In the big data era, university students have gained a more relaxed environment in which to receive and disseminate public opinion information, enabling them to spread their opinions and insights to the Internet more rapidly, thus exacerbating the riskiness of public opinion information dissemination. We constructed CUOPO, the first risk classification dataset of China university online public opinion, and screened out 10,255 representative public opinion texts from a large number of university online public opinion information, including 3,641 risk-free and 6,614 risky texts. These risky texts cover many fields, including 1,755 college livelihood risk texts, 767 campus safety risk texts, 1,395 school order risk texts, 906 university reputation risk texts, and 1,793 advertisement risk texts. The dataset contains various information about each network opinion, including authentic labels, text information, time information, and network information. Through an in-depth study of CUOPO, we found that universities have significant risk issues in the areas of livelihood, safety, teaching order, reputation, and advertisement diversion, which require great attention from university administrators. To validate the effectiveness of the CUOPO, we conduct extensive experiments on the dataset using a series of neural network methods to provide benchmark results for predicting online public opinion risk texts. We expect that CUOPO can provide strong data support for the study of the types of online public opinion risks in colleges and universities and thus play a positive role in promoting the progress of college and university public opinion research. The dataset is available at https://github.com/TianShengLee98/CUOPO-Dataset .
- 16 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 55225-55233
Open AccessChina University Online Public Opinion Risk Dataset
With the widespread popularity of social software and self-media, online public opinion incidents in colleges and universities occur frequently and present a complicated situation. In the big data era, university students have gained a more relaxed environment in which to receive and disseminate public opinion information, enabling them to spread their opinions and insights to the Internet more rapidly, thus exacerbating the riskiness of public opinion information dissemination. We constructed CUOPO, the first risk classification dataset of China university online public opinion, and screened out 10,255 representative public opinion texts from a large number of university online public opinion information, including 3,641 risk-free and 6,614 risky texts. These risky texts cover many fields, including 1,755 college livelihood risk texts, 767 campus safety risk texts, 1,395 school order risk texts, 906 university reputation risk texts, and 1,793 advertisement risk texts. The dataset contains various information about each network opinion, including authentic labels, text information, time information, and network information. Through an in-depth study of CUOPO, we found that universities have significant risk issues in the areas of livelihood, safety, teaching order, reputation, and advertisement diversion, which require great attention from university administrators. To validate the effectiveness of the CUOPO, we conduct extensive experiments on the dataset using a series of neural network methods to provide benchmark results for predicting online public opinion risk texts. We expect that CUOPO can provide strong data support for the study of the types of online public opinion risks in colleges and universities and thus play a positive role in promoting the progress of college and university public opinion research. The dataset is available at https://github.com/TianShengLee98/CUOPO-Dataset .
- Open Access
- 7 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 86963-86976
Chinese Face Dataset for Face Recognition in an Uncontrolled Classroom Environment
Since the position of the classroom surveillance camera is not fixed, the angle of the face captured through the surveillance video is also different. The deep learning-based face verification model has achieved good results in controlled environments, but there is still much room for improvement in the face verification ability in uncontrolled environments. The performance of the model depends not only on the results of the network but also on the quality and diversity of the dataset. The current Asian face dataset in an uncontrolled environment is insufficient; for this reason, this paper constructs a Chinese face dataset (UCEC-Face) in an uncontrolled classroom environment, which is collected by 35 real classroom surveillance videos. The UCEU dataset contains 7395 images of 130 subjects, including 44 males and 86 females. To verify that there is still room for improving the performance of existing face verification models for Asian face verification, we further utilize four models such as OpenFace and ArcFace for face verification, as well as the VGG-Face model for gender, expression, and age recognition on the UCEC-Face. The experimental results show that the UCEC-Face constructed in this paper is more challenging and difficult to verify in face verification tasks because it is closer to the real environment, and the best results obtained on the existing models only reach 69.7%, which is largely below the average accuracy of the identification results of other datasets.
- 7 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 86963-86976
Open AccessChinese Face Dataset for Face Recognition in an Uncontrolled Classroom Environment
Since the position of the classroom surveillance camera is not fixed, the angle of the face captured through the surveillance video is also different. The deep learning-based face verification model has achieved good results in controlled environments, but there is still much room for improvement in the face verification ability in uncontrolled environments. The performance of the model depends not only on the results of the network but also on the quality and diversity of the dataset. The current Asian face dataset in an uncontrolled environment is insufficient; for this reason, this paper constructs a Chinese face dataset (UCEC-Face) in an uncontrolled classroom environment, which is collected by 35 real classroom surveillance videos. The UCEU dataset contains 7395 images of 130 subjects, including 44 males and 86 females. To verify that there is still room for improving the performance of existing face verification models for Asian face verification, we further utilize four models such as OpenFace and ArcFace for face verification, as well as the VGG-Face model for gender, expression, and age recognition on the UCEC-Face. The experimental results show that the UCEC-Face constructed in this paper is more challenging and difficult to verify in face verification tasks because it is closer to the real environment, and the best results obtained on the existing models only reach 69.7%, which is largely below the average accuracy of the identification results of other datasets.