Claim Missing Document
Check
Articles

Found 2 Documents
Search

Facial Movement Recognition Using CNN-BiLSTM in Vowel for Bahasa Indonesia Rahman, Muhammad Daffa Abiyyu; Wicaksono, Alif Aditya; Yuniarno, Eko Mulyanto; Nugroho, Supeno Mardi Susiki
JAREE (Journal on Advanced Research in Electrical Engineering) Vol 8, No 1 (2024): January
Publisher : Department of Electrical Engineering ITS and FORTEI

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.12962/jaree.v8i1.372

Abstract

Speaking is a multimodal phenomenon that has both verbal and non-verbal cues. One of the non-verbal cues in speaking is the facial movement of the subject, which can be used to find the letter being spoken by the subject. Previous research has been done to prove that lip movement can translate to vowels for Bahasa Indonesia, but detecting the whole facial movement is yet to be covered. This research aimed to establish a CNN-BiLSTM model that can learn spoken vowels by reading the subject's facial movements. The CNN-BiLSTM model yielded a 98.66% validation accuracy, with over 94% accuracy for all five vowels. The model is also capable of recognizing whether the subject is currently silent or speaking a vowel with 98.07% accuracy.
Benchmarking deep transfer learning for imbalanced skin cancer classification: Integrating focal loss, explainable AI, and web deployment Aufar, Yazid; Rahman, Muhammad Daffa Abiyyu; Ridhani, M. Fadli
Journal of Soft Computing Exploration Vol. 7 No. 1 (2026): March 2026
Publisher : SHM Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.52465/joscex.v7i1.20

Abstract

Non-melanoma skin cancer (NMSC) classification faces challenges like severe data imbalance and the "black-box" nature of AI, limiting clinical trust. This study benchmarks four pre-trained convolutional models (ConvNeXt-Tiny, EfficientNetV2-S, DenseNet121, MobileNetV3-Large) for the imbalanced multi-class classification of Squamous Cell Carcinoma, Actinic Keratosis, and benign Nevus. Images were preprocessed using morphological hair removal and inpainting. The methodology integrated a 5-fold Stratified Group-KFold cross-validation, Focal Loss to address class imbalance, and Grad-CAM for Explainable AI (XAI) transparency. Results showed ConvNeXt-Tiny achieved the highest and most stable performance with a Balanced Accuracy of 76.98% (± 0.31 standard deviation) and a Macro F1-Score of 0.7513, significantly outperforming the other architectures. Grad-CAM confirmed the model's precise focus on pathological lesion borders. Ultimately, the optimal model was deployed as a real-time Streamlit web application, establishing a robust and practical clinical decision-support system.