Plant disease is a major limiting factor in yield production. Traditionally, biologists rely on manual screening (localisation) of disease symptoms exhibited on plant leaves. However, this is a highly subjective and error-prone analysis. With the advancement of imaging devices, image-based analysis permits high-throughput disease localisation. In this context, state-of-the-art supervised deep learning-based methods have been presented. However, annotation of disease symptoms is time-consuming and is a current bottleneck. To relieve this limitation, we propose a novel unsupervised framework that utilises Cluster weight as well as Group-local Feature-weight learning using iteration in Fuzzy C-Means (CGFFCM) guided via differentiable clustering algorithm for accurate disease localisation. It is to be noted that the proposed framework does not rely on annotated image data. In this paper, we also performed an exhaustive comparison based on different clustering methods to show the efficacy of the proposed framework.