In precision agriculture, the real-time detection of plant leaf diseases using an automated disease detection application has a significant impact on crop production. Researchers have developed several novel models for detecting leaf diseases that exhibit remarkable accuracy and predictive capabilities. However, these novel computer vision models are yet to be implemented as an automated Android-based application that can easily detect and classify different plant leaf diseases. Considering this implementation drawback, this study proposes two novel object detection based computer vision models. These models were developed to detect potato plant leaf diseases using a publicly available dataset of 3,000 images. The first proposed architecture is a customized ‘YOLOv8,’ where ‘depthwise convolution’ is added instead of ‘traditional convolution,’ along with ‘Squeeze-and-Excitation (SE)’ customization. The customized YOLOv8 model outperformed other experimented traditional YOLOv8 and YOLOv5 models with 94.3% accuracy. On the other hand, the proposed customized ViT model, where customization is done by adding the ‘Convolution block,’ ‘Inverted Residual block,’ and ‘MobileViT block,’ outperformed its other variants with 99.33 % accuracy. Although the customized ViT has higher accuracy than the customized YOLOv8 model, the customized YOLOv8 outperforms the customized ViT in terms of its lightweight quality. Finally, after the customization process, the less-weighted, mobile-friendly customized YOLOv8 model was used to implement an Android application. This application will provide rapid relief to farmers by enabling them to detect potato plant leaf diseases.