MaTangDetect: Identification of Potato Pests Using Freezing Layers Ensemble-Based Deep Learning

One of the challenges in the agricultural world is pest attacks that cause crop failure. Potatoes are one of the crops that are often attacked by pests. Early identification of potato pests is crucial because it significantly helps farmers implement appropriate and effective control measures. This not only increases crop yields but also reduces economic losses caused by pest attacks. This study introduces MaTangDetect, a pest detection system that uses an ensemble-based deep learning model with a freezing layer technique. The proposed system uses a freezing layer on the ensemble of DenseNet201, EfficientNetB5, and InceptionV3 models, to improve classification accuracy. By applying the freezing layer technique, the system utilizes previously learned features to the maximum, thereby accelerating the training process and improving overall performance. MaTangDetect was evaluated using various potato pest images and had an accuracy rate of 93%. These results demonstrate the effectiveness of deep learning based on the freezing layer ensemble-based model in identifying potato pest species. In addition, this system offers great potential for application in the field, providing practical and innovative solutions for farmers in pest management in potato cultivation, and contributing to increased yields and reduced economic losses.