Accuracy in plant disease detection is important for sustainable agriculture and preventing crop loss. Conditions are not same in disease prediction; any detection model is expected to perform best in low light areas. In low light areas the common problems faced are degraded image quality, reduced contrast and colour distortion. Though many deep learning algorithms performs good in highlighted areas, they are less accurate in low light environments. We introduce a Convolutional Neural Network (CNN) based Illumination Correction Network (CICN). CICN enhances the low light plant images and it is further processed by a hybrid model including Convolutional Neural Network and Vision Transformers (CNN-ViT). The luminosity of images is enhanced by CICN using an adaptive lighting mapping approach. The proposed method is effective in prediction of sickness under unsatisfactory illumination condition with a maximized accuracy of 92.3% in comparison with Resnet 50, standalone ViT and Efficient Net- B3.