Adaptive weights learning in CNN feature fusion for crime scene investigation image classification
The combination of features from the convolutional layer and the fully connected layer of a convolutional neural network (CNN) provides an effective way to improve the performance of crime scene investigation (CSI) image classification. However, in existing work, as the weights in feature fusion do...
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2021-07-01
|
| Series: | Connection Science |
| Subjects: | |
| Online Access: | http://dx.doi.org/10.1080/09540091.2021.1875987 |
| _version_ | 1797684085514043392 |
|---|---|
| author | Liu Ying Zhang Qian Nan Wang Fu Ping Chiew Tuan Kiang Lim Keng Pang Zhang Heng Chang Chao Lu Lu Guo Jun Ling Nam |
| author_facet | Liu Ying Zhang Qian Nan Wang Fu Ping Chiew Tuan Kiang Lim Keng Pang Zhang Heng Chang Chao Lu Lu Guo Jun Ling Nam |
| author_sort | Liu Ying |
| collection | DOAJ |
| description | The combination of features from the convolutional layer and the fully connected layer of a convolutional neural network (CNN) provides an effective way to improve the performance of crime scene investigation (CSI) image classification. However, in existing work, as the weights in feature fusion do not change after the training phase, it may produce inaccurate image features which affect classification results. To solve this problem, this paper proposes an adaptive feature fusion method based on an auto-encoder to improve classification accuracy. The method includes the following steps: Firstly, the CNN model is trained by transfer learning. Next, the features of the convolution layer and the fully connected layer are extracted respectively. These extracted features are then passed into the auto-encoder for further learning with Softmax normalisation to obtain the adaptive weights for performing final classification. Experiments demonstrated that the proposed method achieves higher CSI image classification performance compared with fix weights feature fusion. |
| first_indexed | 2024-03-12T00:24:30Z |
| format | Article |
| id | doaj.art-9eba7294635842e8aeb5ee02003eb677 |
| institution | Directory Open Access Journal |
| issn | 0954-0091 1360-0494 |
| language | English |
| last_indexed | 2024-03-12T00:24:30Z |
| publishDate | 2021-07-01 |
| publisher | Taylor & Francis Group |
| record_format | Article |
| series | Connection Science |
| spelling | doaj.art-9eba7294635842e8aeb5ee02003eb6772023-09-15T10:47:59ZengTaylor & Francis GroupConnection Science0954-00911360-04942021-07-0133371973410.1080/09540091.2021.18759871875987Adaptive weights learning in CNN feature fusion for crime scene investigation image classificationLiu Ying0Zhang Qian Nan1Wang Fu Ping2Chiew Tuan Kiang3Lim Keng Pang4Zhang Heng Chang5Chao Lu6Lu Guo Jun7Ling Nam8Xi’an University of Posts and TelecommunicationsXi’an University of Posts and TelecommunicationsXi’an University of Posts and TelecommunicationsRekindle Pte LtdXi’an University of Posts and TelecommunicationsXi’an University of Posts and TelecommunicationsXi’an University of Posts and TelecommunicationsFederation UniversitySanta Clara UniversityThe combination of features from the convolutional layer and the fully connected layer of a convolutional neural network (CNN) provides an effective way to improve the performance of crime scene investigation (CSI) image classification. However, in existing work, as the weights in feature fusion do not change after the training phase, it may produce inaccurate image features which affect classification results. To solve this problem, this paper proposes an adaptive feature fusion method based on an auto-encoder to improve classification accuracy. The method includes the following steps: Firstly, the CNN model is trained by transfer learning. Next, the features of the convolution layer and the fully connected layer are extracted respectively. These extracted features are then passed into the auto-encoder for further learning with Softmax normalisation to obtain the adaptive weights for performing final classification. Experiments demonstrated that the proposed method achieves higher CSI image classification performance compared with fix weights feature fusion.http://dx.doi.org/10.1080/09540091.2021.1875987convolutional neural networkauto-encodercrime scene investigation image classificationfeature fusion |
| spellingShingle | Liu Ying Zhang Qian Nan Wang Fu Ping Chiew Tuan Kiang Lim Keng Pang Zhang Heng Chang Chao Lu Lu Guo Jun Ling Nam Adaptive weights learning in CNN feature fusion for crime scene investigation image classification Connection Science convolutional neural network auto-encoder crime scene investigation image classification feature fusion |
| title | Adaptive weights learning in CNN feature fusion for crime scene investigation image classification |
| title_full | Adaptive weights learning in CNN feature fusion for crime scene investigation image classification |
| title_fullStr | Adaptive weights learning in CNN feature fusion for crime scene investigation image classification |
| title_full_unstemmed | Adaptive weights learning in CNN feature fusion for crime scene investigation image classification |
| title_short | Adaptive weights learning in CNN feature fusion for crime scene investigation image classification |
| title_sort | adaptive weights learning in cnn feature fusion for crime scene investigation image classification |
| topic | convolutional neural network auto-encoder crime scene investigation image classification feature fusion |
| url | http://dx.doi.org/10.1080/09540091.2021.1875987 |
| work_keys_str_mv | AT liuying adaptiveweightslearningincnnfeaturefusionforcrimesceneinvestigationimageclassification AT zhangqiannan adaptiveweightslearningincnnfeaturefusionforcrimesceneinvestigationimageclassification AT wangfuping adaptiveweightslearningincnnfeaturefusionforcrimesceneinvestigationimageclassification AT chiewtuankiang adaptiveweightslearningincnnfeaturefusionforcrimesceneinvestigationimageclassification AT limkengpang adaptiveweightslearningincnnfeaturefusionforcrimesceneinvestigationimageclassification AT zhanghengchang adaptiveweightslearningincnnfeaturefusionforcrimesceneinvestigationimageclassification AT chaolu adaptiveweightslearningincnnfeaturefusionforcrimesceneinvestigationimageclassification AT luguojun adaptiveweightslearningincnnfeaturefusionforcrimesceneinvestigationimageclassification AT lingnam adaptiveweightslearningincnnfeaturefusionforcrimesceneinvestigationimageclassification |