Pengenalan Karakter Tulisan Tangan pada Dokumen Berita Acara Bimbingan Skripsi menggunakan Preprocessing dan YOLOv8

Arifah Nur Ainia (a:1:{s:5:"en_US"s:44:"Institut Teknologi Sepuluh Nopember Surabaya"})
Nanik Suciati (Institut Teknologi Sepuluh Nopember)
Hudan Studiawan (Institut Teknologi Sepuluh Nopember)

Jurnal Informatika dan Rekayasa Perangkat Lunak
Vol. 7 No. 2 (2025): September

Handwritten character recognition is one of the challenges in the field of digital image processing, especially in academic documents such as thesis guidance minutes. This study aims to compare the performance of the YOLOv8 model in detecting handwritten characters on two types of datasets, namely original images and preprocessed images. Preprocessing is carried out through the stages of grayscale, CLAHE, Gaussian blur, adaptive thresholding Gaussian, dilation, and erosion. Labels in the preprocessed data are obtained by copying annotations from the original data without adjusting for visual changes. Both datasets were trained using YOLOv8s for 30 epochs. The evaluation results show that the model trained on the original data gives the best results with mAP@0.5 of 0.795 and mAP@0.5:0.95 of 0.606, while the model trained on the preprocessed data only achieves mAP@0.5 of 0.748 and mAP@0.5:0.95 of 0.560.