The growth of essential agricultural crops such as rice, wheat, maize, soybean, and sugarcane are frequently impacted by pest infestations, leading to significant yield reductions due to various insect species. Farmers often face difficulties in accurately identifying and distinguishing these pests, particularly in the early growth stages when many insects exhibit similar features. Early and accurate pest identification is crucial for preventing extensive crop damage. To address this issue, we introduce a novel ensemble model aimed at efficient and precise pest classification, providing a practical solution to assist farmers in overcoming these challenges. The proposed model utilizes two advanced Convolutional Neural Network (CNN) architectures namely EfficientNetB4 and EfficientNetV2-L which were chosen based on their individual effectiveness on a custom dataset of agricultural pests comprising 12 classes. These models achieved accuracy of 91.90% and 92.27%, respectively. Our methodology incorporates an image augmentation and preprocessing pipeline designed to handle variations in insect orientation, such as rotations, flips, and scaling, ensuring the model’s effectiveness in real-world conditions. Furthermore, the preprocessing step resizes the images to pixels \( \textbf{I} \in \mathbb {R}^{256 \times 256 } \) , preparing them for the feature extraction process. The features extracted by both EfficientNet models are combined using an ensemble technique and further refined through dense layers to capture a broader range of visual characteristics for enhanced classification. The final model achieves an overall accuracy of 93.54%, highlighting its robustness and generalization capability. This solution offers considerable potential to help farmers quickly and accurately identify pests, reducing crop losses, improving pest management strategies, and ultimately increasing agricultural productivity.