27 September 2023 in Information Fusion
2 August 2023 in IEEE Transactions on Artificial Intelligence
The fast outbreak of COVID-19 and rapid prolif-eration of its variants have continued to pose a huge challenge to people around the world. Wearing medical masks properly in public and private settings can protect people from COVID-19 which brings a growing demand for automatic detection services of face mask wearing conditions. In this paper, we propose a fine-grained detector called ECA YOLOX-S to identify the wearing status of face masks. Efficient Channel Attention (ECA) is introduced into YOLOX-S to reach a trade-off between effectiveness and efficiency. To demonstrate the performance of our proposed method, a Fine-grained Face Mask (FineFM) dataset is created which covers four classes of mask wearing status. The proposed FineFM dataset has 16,955 annotated images and covers multiple realistic scenarios, which is available at https://github.com/HongliXiao/FineFM . To our knowledge, it has the largest number of improper mask wearing images among all similar datasets for realistic scenes. Experiments conducted on the FineFM dataset demonstrate that ECA YOLOX-S achieves an overall mAP@.50:95 of 86.80% for moderate scenes and an overall mAP@.50:95 of 73.20% for complex scenes, outper-forming its benchmark model. Moreover, experiments conducted on other realistic and simulated datasets indicate the proposed detector has advantages over other methods.
1 June 2022 in Journal of Intelligent & Fuzzy Systems: Applications in Engineering and Technology
Effectively identifying self-admitted technical debt (SATD) from project source code comments helps developers quickly find and repay these debts, thereby reducing its negative impact. Previous studies used techniques based on patterns, text mining, natural language processing, and neural networks to detect SATD. Compared with these above, Convolutional Neural Networks (CNN) have the strong feature extraction ability. Deep network ensembles are demonstrated great potential for the task of sentences classification. In order to boost the performance of CNN-based SATD detecting, we propose a deep neural network ensemble contribute to ensemble learning in a simple yet effective way. Specifically, CNN, CNN-LSTM (convolutional neural network and long short-term memory), and DPCNN (Deep Pyramid Convolutional Neural Networks) are used as individual classifiers to diversify the deep network ensembles. In order to improve the explainability, we introduce attention to measure the contribution of feature words to SATD classification. 62,285 source code comments from 10 projects were used in our experiments. The results show that our approach can effectively reduce misjudgment and detect more SATD, especially for cross-project, so as to greatly improve the detection accuracy.
20 August 2026 in Neural Networks