Image identification of components in circuit diagrams
Small target detection has become an important research project in image processing and deep learning target detection algorithms in recent years. Different from common small target detection objects, this project uses machine learning methods to detect targets for various components in circuit diag...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Master by Coursework |
| Language: | English |
| Published: |
Nanyang Technological University
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/178292 |
| _version_ | 1811687653335105536 |
|---|---|
| author | Zhang, Zhenggege |
| author2 | Tan Eng Leong |
| author_facet | Tan Eng Leong Zhang, Zhenggege |
| author_sort | Zhang, Zhenggege |
| collection | NTU |
| description | Small target detection has become an important research project in image processing and deep learning target detection algorithms in recent years. Different from common small target detection objects, this project uses machine learning methods to detect targets for various components in circuit diagrams. It is different from the traditional identification method of artificially defined target features. It is mainly through training the convolutional neural network model to achieve high-precision feature extraction for complex circuit diagrams containing various components. The YOLOv3 model is You Only Look Once,version 3. It is used as the original framework. A common class of component-connected incomplete circuits is pre-trained on a dataset, and then the model is fine-tuned with complete circuit diagrams. During training, the technology of transfer learning is used to transfer the parameters of the model, which is based on the transfer learning method. In the experimental part, Python is used as the development language, Pytorch is used as the framework of deep learning, and the calculation is performed on the Colab platform in Google. The source of the self-built data set in this dissertation is the image library of Google and Baidu. The quantitative pictures in the data set increase the accuracy of target detection, and the accuracy is gradually improved.The accuracy rate of various classes of recognition can be over 85% or higher. |
| first_indexed | 2024-10-01T05:19:44Z |
| format | Thesis-Master by Coursework |
| id | ntu-10356/178292 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T05:19:44Z |
| publishDate | 2024 |
| publisher | Nanyang Technological University |
| record_format | dspace |
| spelling | ntu-10356/1782922024-06-14T15:43:36Z Image identification of components in circuit diagrams Zhang, Zhenggege Tan Eng Leong School of Electrical and Electronic Engineering EELTan@ntu.edu.sg Engineering Small target detection has become an important research project in image processing and deep learning target detection algorithms in recent years. Different from common small target detection objects, this project uses machine learning methods to detect targets for various components in circuit diagrams. It is different from the traditional identification method of artificially defined target features. It is mainly through training the convolutional neural network model to achieve high-precision feature extraction for complex circuit diagrams containing various components. The YOLOv3 model is You Only Look Once,version 3. It is used as the original framework. A common class of component-connected incomplete circuits is pre-trained on a dataset, and then the model is fine-tuned with complete circuit diagrams. During training, the technology of transfer learning is used to transfer the parameters of the model, which is based on the transfer learning method. In the experimental part, Python is used as the development language, Pytorch is used as the framework of deep learning, and the calculation is performed on the Colab platform in Google. The source of the self-built data set in this dissertation is the image library of Google and Baidu. The quantitative pictures in the data set increase the accuracy of target detection, and the accuracy is gradually improved.The accuracy rate of various classes of recognition can be over 85% or higher. Master's degree 2024-06-13T08:57:20Z 2024-06-13T08:57:20Z 2024 Thesis-Master by Coursework Zhang, Z. (2024). Image identification of components in circuit diagrams. Master's thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/178292 https://hdl.handle.net/10356/178292 en application/pdf Nanyang Technological University |
| spellingShingle | Engineering Zhang, Zhenggege Image identification of components in circuit diagrams |
| title | Image identification of components in circuit diagrams |
| title_full | Image identification of components in circuit diagrams |
| title_fullStr | Image identification of components in circuit diagrams |
| title_full_unstemmed | Image identification of components in circuit diagrams |
| title_short | Image identification of components in circuit diagrams |
| title_sort | image identification of components in circuit diagrams |
| topic | Engineering |
| url | https://hdl.handle.net/10356/178292 |
| work_keys_str_mv | AT zhangzhenggege imageidentificationofcomponentsincircuitdiagrams |