Fang Yang

Dr. Fang Yang

Hebei University, Baoding, Hebei, China
orcid

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

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

Representative Publications

View All Publications
22 March 2023 in Electronics
Verified
Journal Article
Open access
PDF file
FTFan Tang
Fan Tang
Fang YangFang Yang
Fang Yang
XTXianqing Tian
Xianqing Tian
In the research field of small object detection, most object detectors have been successfully used for pedestrian detection, face recognition, lost and found, and automatic driving, among other applications, and have achieved good results. However, when general object detectors encounter challenging low-resolution images from the TinyPerson dataset, they will produce undesirable detection results because of the dense occlusion between people and different body poses. In order to solve these problems, this paper proposes a tiny object detection method TOD-YOLOv7 based on YOLOv7.First, this paper presents a reconstruction of the YOLOv7 network by adding a tiny object detection layer to enhance its detection ability. Then, we use the recursive gated convolution module to realize the interaction with the higher-order space to accelerate the model initialization process and reduce the reasoning time. Secondly, this paper proposes the integration of a coordinate attention mechanism into the YOLOv7 feature extraction network to strengthen the pedestrian object information and weaken the background information.Additionally, we leverage data augmentation techniques to improve the representation learning of the algorithm. The results show that compared with the baseline model YOLOv7, the detection accuracy of this model on the TinyPerson dataset is improved from 7.1% to 9.5%, and the detection speed reaches 208 frames per second (FPS). The algorithm of this paper is shown to achieve better detection results for tiny object detection.

11 February 2023 in Geomatics, Natural Hazards and Risk
Verified
Journal Article
Open access
YLYongfa Li
Yongfa Li
XZXiaoqing Zuo
Xiaoqing Zuo
Fang YangFang Yang
Fang Yang
Jinwei BuJinwei Bu
Jinwei Bu
WWWenhao Wu
Wenhao Wu
XLXinyu Liu
Xinyu Liu
InSAR technology provides a powerful tool for detecting large-scale surface deformation. In particular, the newly developed DS-InSAR method fused PS points has obvious advantages in monitoring bare land and vegetation covered areas, but at present, there is a lack of effectiveness evaluation of this method. Therefore, this paper takes Hongta District of Yuxi City as an example. The 29 scenes sentinel-1A data from January 2019 to December 2019 were processed and analyzed using the DS-InSAR method fused PS points. The research results show that the position and deformation trend of the inversion results of the three methods are highly consistent, the correlation between DS-InSAR method fused PS points and PS- InSAR and SBAS-InSAR methods is 0.9473 and 0.8583, respectively. But the spatial density of the measurement points (MPs) obtained by the DS-InSAR method of PS points is 11 times and 4 times that of the PS-InSAR method and SBAS-InSAR method respectively. It clearly shows that DS-InSAR method fused PS points has greater advantages than PS-InSAR and SBAS-InSAR methods in deformation mapping displacement, because the MPs density is higher. It is conducive to the detailed analysis of the spatio-temporal characteristics and deformation mechanism of deformation.

19 February 2023 in Sensors
Verified
Journal Article
Open access
PDF file
CRCai Ran
Cai Ran
XLXinfu Li
Xinfu Li
Fang YangFang Yang
Fang Yang
The proliferation of deep learning has propelled image inpainting to an important research field. Although the current image inpainting model has made remarkable achievements, the two-stage image inpainting method is easy to produce structural errors in the rough stage because of insufficient treatment of the rough inpainting stage. To address this problem, we propose a multi-step structured image inpainting model combining attention mechanisms. Different from the previous two-stage inpainting model, we divide the damaged area into four sub-areas, calculate the priority of each area according to the priority, specify the inpainting order, and complete the rough inpainting stage several times. The stability of the model is enhanced by the multi-step method. The structural attention mechanism strengthens the expression of structural features and improves the quality of structure and contour reconstruction. Experimental evaluation of benchmark data sets shows that our method effectively reduces structural errors and improves the effect of image inpainting.

15 September 2022 in IEEE Access
Verified
Journal Article
Open access
XGXiang Gao
Xiang Gao
Fang YangFang Yang
Fang Yang
TCTian Chen
Tian Chen
JSJianhui Si
Jianhui Si
To solve the component segmentation problem caused by the sticking and overlapping of parts in incoherent handwritten calligraphy characters, we propose a Chinese character part segmentation method based on Faster RCNN. The method utilizes the advantages of Faster RCNN on multi-scale and small targets to solve difficult problems in component segmentation. The hierarchical features of the components were used in our proposed method to identify each layer of the Chinese character structure to obtain the components. Qualitative and quantitative calculations were used to test the segmentation effect of the proposed method. The experimental results demonstrate the accurate segmentation effectiveness of our method for adhering and overlapping components. In addition, these components could be retrieved accurately in the retrieval system, and the mean Average Precision of the top 30 retrieval results reached 95.7%. A better retrieval accuracy reflects a better segmentation effect from the side, which proves the effectiveness of the proposed method.

14 July 2022 in Electronics
Verified
Journal Article
Open access
PDF file
YSYao Song
Yao Song
Fang YangFang Yang
Fang Yang
TLTe Li
Te Li
The calligraphic works of particular calligraphers often contain only a limited number of characters, rather than the full set of Chinese characters required for typography, which does not meet practical needs. There is therefore a need to develop a complete set of calligraphic characters for calligraphers. Most of the recently popular methods for generating calligraphic characters are based on deep learning, using an end-to-end approach to generate the target image. Deep learning-based methods usually suffer from unsuccessful conversion of stroke structures when the printed font differs significantly from the target font structure. In this paper, we propose an involution-based calligraphic character generation model, which can realize the conversion from printed fonts to target calligraphic fonts. We improve the Pix2Pix model by using a new neural operator, involution, which focuses more on spatial feature processing and can better handle the relationship between strokes than the models using only convolution, so that the generated calligraphic characters have an accurate stroke structure. A self-attentive module and a residual block are also added to increase the depth of the network to improve the feature processing capability of the model. We evaluated our method and some baseline methods using the same dataset, and the experimental results demonstrate that our model is superior in both visual and quantitative evaluation.

4 August 2022 in Applied Sciences
Verified
Journal Article
Open access
PDF file
JSJingya Song
Jingya Song
QSqingxuan shi
qingxuan shi
LYLi Yihang
Li Yihang
Fang YangFang Yang
Fang Yang
Human parsing is a fine-grained human semantic segmentation task in the field of computer vision. Due to the challenges of occlusion, diverse poses and a similar appearance of different body parts and clothing, human parsing requires more attention to learn context information. Based on this observation, we enhance the learning of global and local information to obtain more accurate human parsing results. In this paper, we introduce a Global Transformer Module (GTM) via a self-attention mechanism to capture long-range dependencies for effectively extracting context information. Moreover, we design a Detailed Feature Enhancement (DFE) architecture to exploit spatial semantics for small targets. The low-level visual features from CNN intermediate layers are enhanced by using channel and spatial attention. In addition, we adopt an edge detection module to refine the prediction. We conducted extensive experiments on three datasets (i.e., LIP, ATR, and Fashion Clothing) to show the effectiveness of our method, which achieves 54.55% mIoU on the LIP dataset, 80.26% on the average F-1 score on the ATR dataset and 55.19% on the average F-1 score on the Fashion Clothing dataset.