Skripsi

IMAGE CAPTIONING PADA CITRA SERVIKS UNTUK LAPORAN MEDIS BERBASIS MODEL DEEP LEARNING

---

Image captioning is a task in the fields of computer vision (CV) and natural language processing (NLP) that aims to generate textual descriptions from an image. In this study, various combinations of encoder–decoder architectures were designed and evaluated to improve captioning performance on cervical medical images from the International Agency for Research on Cancer (IARC). The encoders used include EfficientNet, ResNet, and YOLOv11, while the decoders consist of Transformer and LSTM, resulting in six model combinations being tested. Each model was trained using various hyperparameter configurations to achieve optimal performance. Evaluation was conducted using the Bilingual Evaluation Understudy (BLEU) and Metric for Evaluation of Translation with Explicit Ordering (METEOR) metrics to measure the alignment between generated text and ground truth. The experimental results show that the ResNet–LSTM combination achieved the best performance, with BLEU 1–4 scores of 0.7343, 0.6762, 0.6360, and 0.6029, and a METEOR score of 0.7603. Furthermore, the best-performing model was optimized using the Low-Rank Adaptation (LoRA) method. The implementation of LoRA successfully improved parameter efficiency by reducing the number of trainable parameters from 46,511,816 to just 6,669,000. However, the evaluation results indicate a decline in performance after applying LoRA, suggesting a trade-off between parameter efficiency and the quality of generated outputs.

---

Availability

Detail Information

Series Title

-

Call Number

T2001052026

Publisher

Indralaya : Prodi Sistem Komputer, Fakultas Ilmu Komputer Universitas Sriwijaya., 2026

Collation

xiv, 99 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

006.307

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Prodi Sistem Komputer
Model Deep Learning

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment