Xiaoming Zhang

Xiaoming Zhang

Auto-generated profile, based on publicly available information

Xiaoming Zhang

Department of Computer and Science, College of Information Engineering, Beijing Institute of Petrochemical Technology, Beijing, China

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

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

Representative Publications

View All Publications
8 August 2018 in IOP Conference Series: Materials Science and Engineering
Verified
Conference Paper
Open access
PDF file
XZXiaoming Zhang
Xiaoming Zhang
YLYongqiu Lan
Yongqiu Lan
YLYong Liang
Yong Liang

8 October 2021 in Security and Communication Networks
Verified
Journal Article
Open access
PDF file
RYRuiqing Yan
Ruiqing Yan
LSLanchang Sun
Lanchang Sun
Fang WangFang Wang
Fang Wang
XZXiaoming Zhang
Xiaoming Zhang
Recently, pretrained language models, such as Bert and XLNet, have rapidly advanced the state of the art on many NLP tasks. They can model implicit semantic information between words in the text. However, it is solely at the token level without considering the background knowledge. Intuitively, background knowledge influences the efficacy of text understanding. Inspired by this, we focus on improving model pretraining by leveraging external knowledge. Different from recent research that optimizes pretraining models by knowledge masking strategies, we propose a simple but general method to transfer explicit knowledge with pretraining. To be specific, we first match knowledge facts from a knowledge base (KB) and then add a knowledge injunction layer to a transformer directly without changing its architecture. This study seeks to find the direct impact of explicit knowledge on model pretraining. We conduct experiments on 7 datasets using 5 knowledge bases in different downstream tasks. Our investigation reveals promising results in all the tasks. The experiment also verifies that domain-specific knowledge is superior to open-domain knowledge in domain-specific task, and different knowledge bases have different performances in different tasks.

25 October 2024
Verified
Conference Paper
Open access
XZXiaoming Zhang
Xiaoming Zhang
YGYufan Guan
Yufan Guan
QLQuanlong Li
Quanlong Li
In unmanned intelligent refueling systems, complex scenarios such as dirty vehicle bodies, low light, and improper angles often directly affect the detection of fuel tank lids. This paper proposes a comprehensive improved model, BES-YOLO, based on the YOLOv8 framework. First, three instances of the BiFPN multi-layer feature extraction network were added to the architecture to enhance the detection accuracy of small objects. Second, an effective attention mechanism module (EMA) was introduced, adding EMA modules to each output layer of the detection head to strengthen feature representation. Third, the SIOU loss function was adopted to optimize the distance and relative angle between predicted and ground truth boxes. Experimental results show that compared to YOLOv8n, the BES-YOLO model achieved an accuracy of 98.2%, mAP@50 of 99.2%, and mAP@50:95 of 71.9% on the self-built dataset, with increases of 0.2%, 0.2%, and 1.1%, respectively. Testing on the VOC dataset yielded results of 75.6%, 74.0%, and 51.8%, representing increases of 5.4%, 1.6%, and 1.2%, respectively, demonstrating significant detection performance.