PRATAMA, WAHYU PUTRA (2026) Klasifikasi dan Deteksi Katarak Menggunakan Model Ensemble Learning (ResNet-EfficientNet) dengan Teknik Layer-CAM. Undergraduate thesis, UPN Veteran Jawa Timur.
|
Text (Cover)
Cover.pdf Download (2MB) |
|
|
Text (BAB 1)
BUKU SKRIPSI WAHYU FINAL-BAB 1.pdf Download (150kB) |
|
|
Text (BAB 2)
BUKU SKRIPSI WAHYU FINAL-BAB 2.pdf Restricted to Repository staff only until 13 November 2028. Download (481kB) |
|
|
Text (BAB 3)
BUKU SKRIPSI WAHYU FINAL-BAB 3.pdf Restricted to Repository staff only until 13 November 2028. Download (640kB) |
|
|
Text (BAB 4)
BUKU SKRIPSI WAHYU FINAL-BAB 4.pdf Restricted to Repository staff only until 13 November 2028. Download (4MB) |
|
|
Text (BAB 5)
BUKU SKRIPSI WAHYU FINAL-BAB 5.pdf Download (175kB) |
|
|
Text (DAFTAR PUSTAKA)
BUKU SKRIPSI WAHYU FINAL-DAFTAR PUSTAKA.pdf Download (90kB) |
|
|
Text (LAMPIRAN)
BUKU SKRIPSI WAHYU FINAL-LAMPIRAN.pdf Restricted to Repository staff only until 13 November 2028. Download (85kB) |
Abstract
Cataracts are a leading cause of blindness, yet they are often detected too late due to limitations in the availability of ophthalmologists and the subjectivity of manual examinations. This study examines ways to improve the accuracy and interpretability of deep learning-based cataract detection models, particularly addressing the instability of single CNN models. This study aims to develop a cataract classification and detection model using an ensemble learning approach that combines the ResNet and EfficientNet architectures with the stacking technique. The dataset used in this study consists of 2,421 eye images divided into three classes: Normal, Immature Cataract, and Mature Cataract. This combination of models was chosen to leverage ResNet’s strength in training stability and EfficientNet’s strength in feature extraction efficiency. Additionally, this study applies Layer-CAM as an Explainable Artificial Intelligence (XAI) method to enhance model interpretability through the visualization of image regions that influence predictions. The results show that ResNet-18 and EfficientNet-B0 are the best base learners, with accuracies of 92.62% and 94.01%, respectively. In the ensemble stage, the SVM meta-learner achieved the highest performance with an accuracy of 99.18%, precision of 99%, recall of 99%, and an F1-score of 99%, demonstrating the ability to accurately distinguish between the three classes: Normal, Immature Cataract, and Mature Cataract. The Layer-CAM visualization also shows that the model consistently focuses on the lens area of the eye as the basis for decision-making, making the prediction results easier to understand. This research is expected to support the development of an accurate, efficient, and interpretable AI-based cataract screening system.
| Item Type: | Thesis (Undergraduate) | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Contributors: |
|
||||||||||||
| Subjects: | Q Science > QA Mathematics > QA76.6 Computer Programming Q Science > QA Mathematics > QA76.76.E95 Expert Systems Q Science > QA Mathematics > QA76.87 Neural computers |
||||||||||||
| Divisions: | Faculty of Computer Science > Departemen of Data Science | ||||||||||||
| Depositing User: | WPP WAHYU PUTRA PRATAMA | ||||||||||||
| Date Deposited: | 13 Nov 2026 04:17 | ||||||||||||
| Last Modified: | 13 Nov 2026 06:07 | ||||||||||||
| URI: | https://repository.upnjatim.ac.id/id/eprint/61347 |
Actions (login required)
![]() |
View Item |
