Cassava (Manihot esculenta) is an essential crop in Vietnam’s agricultural sector, contributing significantly to food security and economic development. However, monitoring cassava cultivation remains challenging due to the limitations of manual field surveys and the lack of scalable solutions. This study proposes a deep learning-based approach for segmenting cassava fields using high-resolution satellite imagery collected over Tan Hoi commune, Tan Chau district, Vietnam. A labeled dataset was constructed from satellite images to train and evaluate segmentation models, with techniques applied to address class imbalance and improve generalization. The results show that the proposed method achieves accurate and robust segmentation, demonstrating its potential for enhancing cassava monitoring and supporting precision agriculture practices in rural regions.