JutePestDetect: An intelligent approach for jute pest identification using fine-tuned transfer learning

In certain Asian countries, Jute is one of the primary sources income and Gross Domestic Product (GDP) for agricultural sector. Like many other crops, prone to pest infestations, its identification typically made visually in countries like Bangladesh, India, Myanmar, China. addition, this method time-consuming, challenging, somewhat imprecise, which poses a substantial financial risk. To address issue, study proposes high-performing resilient transfer learning (TL) based JutePestDetect model identify jute pests at early stage. Firstly, we prepared dataset containing 17 classes around 380 photos per class, were evaluated after manual automatic pre-processing cleaning, such as background removal resizing. Subsequently, five prominent pre-trained models—DenseNet201, InceptionV3, MobileNetV2, VGG19, ResNet50—were selected from previous design model. Each was revised by replacing classification layer with global average pooling incorporating dropout regularization. evaluate models' performance, various metrics precision, recall, F1 score, ROC curve, confusion matrix employed. These analyses provided additional insights determining efficacy models. Among them, customized regularized DenseNet201-based proposed outperformed others, achieving an impressive accuracy 99%. As result, our strategy offer enhanced approach case Jute, can significantly benefit farmers worldwide.