Cassava (Manihot esculenta) is a strategic agricultural commodity whose productivity is frequently threatened by leaf diseases such as bacterial blight, brown streak, green mottle, and mosaic disease. Manual identification by humans tends to be subjective, time-consuming, and prone to error. This study aims to develop an automatic and intelligent Cassava leaf disease classification system based on Deep Learning that is both accurate and efficient. To overcome the computational burden of conventional models and address real-world data challenges, such as class imbalance and lighting variations, this research proposes the use of the EfficientNet architecture combined with the Transfer Learning method. The model utilizes pre-trained weights from ImageNet to accelerate convergence and optimize visual feature extraction. Experimental results on the Cassava leaf image dataset show that the proposed model successfully achieved an accuracy rate of 81%. These findings demonstrate that the EfficientNet approach provides objective predictions with high computational efficiency. This research has significant potential for implementation in portable devices as an early detection tool for farmers, supporting rapid mitigation actions and maintaining global food security stability