The Advantage of Transfer Learning with Pre-Trained Model in CNN Towards Ct-Scan Classification
Medical image classification plays significant role in the process of medical decisions making, especially during the difficult period of the pandemic. One method being considered good at such classification is Convolutional Neural Network, in which we use pre-trained model approach with transfer le...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | UMS Journal (OJS) |
| Language: | eng |
| Published: |
Department of Informatics, Universitas Muhammadiyah Surakarta, Indonesia
2023
|
| Subjects: | |
| Online Access: | https://journals.ums.ac.id/index.php/khif/article/view/19872 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1805342468597612544 |
|---|---|
| author | Pardede, Jasman S. Purohita, Adwityo |
| author_facet | Pardede, Jasman S. Purohita, Adwityo |
| author_sort | Pardede, Jasman |
| collection | OJS |
| description | Medical image classification plays significant role in the process of medical decisions making, especially during the difficult period of the pandemic. One method being considered good at such classification is Convolutional Neural Network, in which we use pre-trained model approach with transfer learning since the limitation of medical images may require optimal effort. Through this pre-trained model with transfer learning, the objective is to maximize the accuracy of classification and to push forward the training session throughout the comparison of both transfer learning and non-transfer learning based pre-train models. The first type provides average accuracy of 0.84 with approximate training time 0.54 hour while the latter shows the average result of accuracy as 0.74 with average training time 0.58 hour. As the result, the optimizations are 1.13x for accuracy and 1.1x for training time. EfficientNetV2 is one pre-trained model selected for this project, being exposed to both transfer learning and non-transfer learning approach systems. The transfer learning version provides the superior accuracy as 0.88 and training time as 31 minutes - 50 seconds, showing the accuracy of 0.94 on validation and 0.88 on testing |
| format | UMS Journal (OJS) |
| id | oai:ojs2.journals.ums.ac.id:article-19872 |
| institution | Universitas Muhammadiyah Surakarta |
| language | eng |
| publishDate | 2023 |
| publisher | Department of Informatics, Universitas Muhammadiyah Surakarta, Indonesia |
| record_format | ojs |
| spelling | oai:ojs2.journals.ums.ac.id:article-19872 The Advantage of Transfer Learning with Pre-Trained Model in CNN Towards Ct-Scan Classification Pardede, Jasman S. Purohita, Adwityo CT-Scan; CNN; Classification; EfficientNetV2; Transfer Learning Medical image classification plays significant role in the process of medical decisions making, especially during the difficult period of the pandemic. One method being considered good at such classification is Convolutional Neural Network, in which we use pre-trained model approach with transfer learning since the limitation of medical images may require optimal effort. Through this pre-trained model with transfer learning, the objective is to maximize the accuracy of classification and to push forward the training session throughout the comparison of both transfer learning and non-transfer learning based pre-train models. The first type provides average accuracy of 0.84 with approximate training time 0.54 hour while the latter shows the average result of accuracy as 0.74 with average training time 0.58 hour. As the result, the optimizations are 1.13x for accuracy and 1.1x for training time. EfficientNetV2 is one pre-trained model selected for this project, being exposed to both transfer learning and non-transfer learning approach systems. The transfer learning version provides the superior accuracy as 0.88 and training time as 31 minutes - 50 seconds, showing the accuracy of 0.94 on validation and 0.88 on testing Department of Informatics, Universitas Muhammadiyah Surakarta, Indonesia 2023-10-29 info:eu-repo/semantics/article info:eu-repo/semantics/publishedVersion application/pdf https://journals.ums.ac.id/index.php/khif/article/view/19872 10.23917/khif.v9i2.19872 Khazanah Informatika : Jurnal Ilmu Komputer dan Informatika; Vol. 9 No. 2 October 2023; 155-161 Khazanah Informatika; Vol. 9 No. 2 October 2023; 155-161 2477-698X 2621-038X eng https://journals.ums.ac.id/index.php/khif/article/view/19872/8763 Copyright (c) 2023 Jasman Pardede, Adwityo S. Purohita https://creativecommons.org/licenses/by/4.0 |
| spellingShingle | CT-Scan; CNN; Classification; EfficientNetV2; Transfer Learning Pardede, Jasman S. Purohita, Adwityo The Advantage of Transfer Learning with Pre-Trained Model in CNN Towards Ct-Scan Classification |
| title | The Advantage of Transfer Learning with Pre-Trained Model in CNN Towards Ct-Scan Classification |
| title_full | The Advantage of Transfer Learning with Pre-Trained Model in CNN Towards Ct-Scan Classification |
| title_fullStr | The Advantage of Transfer Learning with Pre-Trained Model in CNN Towards Ct-Scan Classification |
| title_full_unstemmed | The Advantage of Transfer Learning with Pre-Trained Model in CNN Towards Ct-Scan Classification |
| title_short | The Advantage of Transfer Learning with Pre-Trained Model in CNN Towards Ct-Scan Classification |
| title_sort | advantage of transfer learning with pre trained model in cnn towards ct scan classification |
| topic | CT-Scan; CNN; Classification; EfficientNetV2; Transfer Learning |
| topic_facet | CT-Scan; CNN; Classification; EfficientNetV2; Transfer Learning |
| url | https://journals.ums.ac.id/index.php/khif/article/view/19872 |
| work_keys_str_mv | AT pardedejasman theadvantageoftransferlearningwithpretrainedmodelincnntowardsctscanclassification AT spurohitaadwityo theadvantageoftransferlearningwithpretrainedmodelincnntowardsctscanclassification AT pardedejasman advantageoftransferlearningwithpretrainedmodelincnntowardsctscanclassification AT spurohitaadwityo advantageoftransferlearningwithpretrainedmodelincnntowardsctscanclassification |