scispace - formally typeset
Search or ask a question
Author

Fangfang Chen

Bio: Fangfang Chen is an academic researcher from Ocean University of China. The author has contributed to research in topics: Feature extraction & Semantic similarity. The author has co-authored 3 publications.

Papers
More filters
Proceedings ArticleDOI
23 Jul 2021
TL;DR: In this paper, a lightweight model AMSSD based on SSD reconstruction was proposed to further improve the accuracy and efficiency of facial expression recognition, which achieved 0.733 mAP on the FER2013 facial expression data set, which fully verified the effectiveness of the method.
Abstract: Facial expression recognition is a technology that uses computers to understand human emotions to provide a good human-computer interaction experience. To further improve the accuracy and efficiency of facial expression recognition, we propose a lightweight model AMSSD based on SSD reconstruction. In this paper, we use the MobileNetV2 lightweight network to replace the SSD base net to reduce the amount of parameters. And use CBAM to improve feature extraction capabilities from the two dimensions of channel and space. This method reduces the number of parameters and improves the efficiency and accuracy of recognition. The AMSDD model achieves 0.733 mAP on the FER2013 facial expression data set, which fully verified the effectiveness of the method.
Proceedings ArticleDOI
23 Jul 2021
TL;DR: Wang et al. as mentioned in this paper employed the siamese BiLSTM network as sequential inference model to extract the global features and the coattention mechanism is employed to generate attention weight between the text features, which will collect local inference over sequences.
Abstract: Text similarity measurement is a basic task in natural language processing and widely used in information retrieval, automatic question answering, machine translation, etc. Because most traditional statistical-based methods for text similarity measurement cannot efficiently extract the semantic information of the text, we propose a BiLSTM-SECapsNet hybrid model based on BiLSTM, CapsNet, and SENet for text similarity measurement. We employed the siamese BiLSTM network as sequential inference models to extract the global features. The coattention mechanism is employed to generate attention weight between the text features, which will collect local inference over sequences. The CapsNet network is also introduced to catch the local features of the text. And then the SENet network is used to automatically calibrate the importance of each local feature to obtain the local feature matrix. After that, the feature matrix is fused again and the BiLSTM network is used to extract the context information to obtain the similarity matrix of the two texts. At last, the semantic similarity of the text is measured through the fusion, pooling, and fully connected layer. The experimental results based on the Quora Questions Pairs data set show that the accuracy of the method is 87.31, and the F-measure is 87.35. Compared with other networks, the effectiveness of the method has been improved.
Proceedings ArticleDOI
23 Jul 2021
TL;DR: In this paper, an improved YOLOv3-45k method was proposed to further improve the recognition effect, which reduced the number of residual modules in feature extraction network to improve the detection speed.
Abstract: Image recognition of refrigerator items has become one of the hotspots in computer vision with the widespread usage of an intelligent refrigerator. There are some problems, such as slow detection speed, missed detection, false detection, when applying a convolutional neural network for the recognition of refrigerator image. To further improve the recognition effect, we proposed an improved YOLOv3-45k method. Firstly, the number of residual modules in feature extraction network is reduced, so that to improve the detection speed. Secondly, the $\mathbf{K}-\mathbf{means}++$ clustering algorithm is used to obtain the anchors of refrigerator items, which is adjusted reasonably by manual method, and then corresponding to their respective feature layers to improve the detection accuracy. The experimental results show that the proposed algorithm can detect 57 images per second on the refrigerator item data set, and the mAP is 84.40%. Compared with the original YOLOv3 algorithm, the proposed YOLOv3-45k method has better detection accuracy and detection speed in the refrigerator item data set.