Penerapan Vision Transformer Untuk Klasifkasi Sampah Rumah Tangga

Abstract

The increasing volume of Household waste requires an accurate  and efficient automatic waste sorting system. This study aims to apply  Vision Transformer (ViT) for image-based household waste clasification. The dataset was divided inti training and validation sets and prepared to match the Vision Transformer archtecture. The ViT-Base Patch16-224 model was trained using the AdamW optimizer with a learning rate of 0.0002, batch size of 16, and 15 training epoch. Model performence was evaluated using accuracy, precision, recall, F1-score, and confusion matrix. Experimental results show that the proposed model achieved an overall accuracy of 95%. The inorganic class obtained a precision of 0.9,  recall of 0.96,  and F1-score of 0.95, while the organic class achived a precision of 0.94, recall of  0.93, F1-score of 0.94. these result indicate that self-attention mechanism in Vision Transformer effectively extracts global visual features and improves clasification stability. Therefore, Vision Transformer dermonstrates strong potential for implementasi in intelligent automatic waste sorting systems.