Claim Missing Document
Check
Articles

Tuning the Alpha Hyperparameter in the Multires U-Net Architecture for Segmentation of Bali Pendet Dance Images Darmawan Bakti, Lalu; Nasirudin Karim, Muh; Imran, Bahtiar
Jurnal Media Elektrik Vol. 23 No. 2 (2026): MEDIA ELEKTRIK
Publisher : Jurusan Pendidikan Teknik Elektro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59562/metrik.v23i2.11944

Abstract

Bali’s traditional Pendet dance represents an important cultural heritage that requires preservation. To support dance recognition, this study applied semantic segmentation to Pendet dance images using the Multires U-Net architecture with alpha hyperparameter tuning. Specifically, three optimization methods Particle Swarm Optimization (PSO), Grid Search, and Random Search were evaluated using the Jaccard Index, Dice Coefficient, and Mean Squared Error (MSE). The results demonstrate that Grid Search produced the optimal alpha value of 1.45, achieving average Jaccard and Dice scores of 98.5002 and 99.2439, respectively. These results outperform previous research (98.4746; 99.2309), PSO (98.4883; 99.2378), and Random Search (98.4837; 99.2352). For MSE, the prior study reported the best score of 7.608E-04, followed by Grid Search (7.659E-04), PSO (7.663E-04), and Random Search (7.765E-04). These findings highlight the effectiveness of Grid Search in optimizing the alpha hyperparameter for the Multires U-Net architecture and demonstrate a significant performance improvement compared to earlier studies.
COLONOSCOPIC POLYP SEGMENTATION USING SEGFORMER-B0 WITH A DICE-BCE HYBRID LOSS Ahmad Yani; San Sudirman; M. Zulpahmi; Emi Suryadi; Bahtiar Imran
Jurnal Kecerdasan Buatan dan Teknologi Informasi Vol. 5 No. 2 (2026): May 2026
Publisher : Ninety Media Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.69916/jkbti.v5i2.476

Abstract

Colorectal cancer is one of the leading causes of cancer-related deaths worldwide, with most cases originating from early lesions such as colon polyps. Early detection through colonoscopy is essential to reduce mortality rates; however, accurate polyp identification remains challenging due to variations in shape, size, texture, and illumination conditions. This study aims to implement and evaluate the SegFormer-B0 architecture combined with a Dice-BCE hybrid loss function for polyp segmentation in colonoscopy images. The study utilized the public Kvasir-SEG dataset consisting of 1,000 colonoscopy images with pixel-level annotations. The dataset was divided into 80% training data and 20% validation data. Image preprocessing included resizing to 256×256 pixels and normalization using ImageNet statistics. The model was trained for 25 epochs using the AdamW optimizer with a learning rate of 1×10⁻⁴. Performance evaluation was conducted using Dice Coefficient, Intersection over Union (IoU), Sensitivity, and Specificity metrics. The experimental results demonstrated that the proposed model achieved a Dice Coefficient of 89.92%, Mean IoU of 81.90%, Sensitivity of 89.12%, and Specificity of 98.51%. The training process also showed stable convergence, supported by a training loss of 7.53% and validation loss of 23.30%. The findings indicate that the integration of SegFormer-B0 with the Dice-BCE hybrid loss effectively improves segmentation accuracy and stability while addressing class imbalance issues in colonoscopy images. Therefore, the proposed approach has strong potential to support computer-aided diagnosis systems for colorectal cancer screening.
Co-Authors AA Sudharmawan, AA Abba Suganda Girsang, Abba Suganda Ahmad Yani ahmad yani Akbar, Ardiyallah Akhmad Muzakka Alfian Hidayat Amirudin Kalbuadi Atika Zahra Nirmala Baihaki, Makmun Baiq Nonik Ria Riska Baiq Nonik Ria Riska Darmawan Bakti, Lalu Diki Hananta Firdaus Efendi, Muhamad Masjun Erfan Wahyudi erniwati, surni Fachrul Kurniawan Febri, Elin Febriani Giardi, Muh Hamzah Andung Hambali Hambali Hambali Hambali Hamim, Lutfi Hasan Basri Hidayatullah, Beni Ari Karim, Muh Nasirudin Karim, Muh. Nasirudin Karina Nurwijayanti Karya Gunawan Karya Gunawan Lalu Darmawan Bakti Lalu Darmawan Bakti, Lalu Darmawan Lalu Delsi Samsumar, M.Eng. M Zulpahmi M. Zulpahmi M. Zulpahmi Mahayadi, Mahayadi Makmun Baihaki Marroh, Zahrotul Isti’anah Maspaeni Maspaeni Moch Arief Soeleman, Moch Arief Muahidin, Zumratul Muh. Akshar Muhammad Masjun Efendi Muhammad Rijal Alfian Muhammad Zohri Mutaqin, Zaenul Muttaqin, Athaur Muzakka, Akhmad Nasirudin Karim, Muh Ndang, Rijalul Mujahidin Nining Putri Ningsih Nunung Rahmania Nurkholis, Lalu Moh. Pratama, Rifqy Hamdani Purnamasidi, Hanis Purwanto Purwanto Ramdan, Hendri Ricardus Anggi Pramunendar Riska, Baiq Nonik Ria Rosida, Sri Rudi Muslim Rudi Muslim Salman Salman Salman Salman Salman San Sudirman Saputra, Dede Haris Satriawan, Andre Selamet Riadi Selamet Riadi Soeleman, Moh. Arief Sriasih, Sriasih Subektiningsih Subektiningsih Subki, Ahmad Suharjito Suharjito, Suharjito Suhartono Supardianto Supardianto Surni Erniwati Suryadi, Emi Tahrir, Muhammad wahyuni, wenti ayu Zaeniah Zaeniah Zaeniah Zaeniah Zaenudin Zaenudin Zaenudin Zaenudin Zaenudin Zaenudin Zaenudin Zaenudin Zaenudin Zahroni, Teguh Rizali Zenuddin, Z Zulpahmi, M Zulpahmi, M. Zulpan Hadi Zulpan Hadi