Ubiquitously promoting digital system emulates the human visual capabilities to get forward every manual step for fully automated system. Agriculture is one of the most necessary and fruitful domains for computer scientists to meet the satisfactory requirement of global and sustainable development. The meaning of sustainable development is pasteurized due to natural resources. The pollutant air, global warming badly impact of plants leaves. This article attempts to present a machine learning based leaf disease modeling framework. Plant diseases have contiguous impact and spread sharply; therefore, it is important to have an early detection of the disease. Considering the importance and badly impacting issues in leaf disease detection, we introduce Separable CNN Vision Transformer (SCN2Vit) model which utilizes a muti-headed vision transformer learnt from mobileNetv2 in transfer learning mode. The learnt feature to enhance the performance is trained from YOLOv7, a faster and real-time most accurate object detector. For optimizing the network Nesterov-Accelerated Adaptive Momentum (NADM) is used and ensured that SCN2Vit-YOLOv7 meets the challenges satisfactorily. The experiments were performed to test the model on publicly available collected from Makerere AI laboratory NaCRR institute.