With the technological advancement in agricultural industry 4.0 emerged has the significant benefits to recent practices in agriculture. However, crop yields are infringed by the similar types of diseases that are misclassified with the same textural information. Moreover, the agricultural industry has experienced a decline in crop productivity due to pests and diseases. Therefore, to address this concern disease identification in early stage has opened up the new dimension in the present practices. Recent developments of transformer in deep learning have revolutionized and opened new dimensions in this field. Unlike, the traditional deep learning model focuses on extracting only the local structural features of image. To overcome this challenge, a novel hybrid method, AgriTransformer, is developed where the CNN captures the local, and transformer captures the global information of images. In the transformer network, leveraging Multi-Head Self-Attention (MSA) mechanism is used to capture the long-term dependencies. Furthermore, the proposed model is tested on the different publicly available and real-field dataset. The efficacy of the model is compared with the other state-of-the-art methods transformer models and achieved the accuracy of 98.50% .