Parkinson’s disease, impacting millions globally, is a progressive degeneration of the nervous system, characterized by shaking sensations, stiffness, and impaired motor skills and handwriting. These symptoms influence handwriting capabilities from the onset. Early detection is crucial for enhancing the way of living in individuals with PD. This study utilizes Meander HandPD dataset having handwriting samples from healthy and PD patients. This study contributes by Efficient-NetB4 and an ensemble model EfficientNetB4 + SVM for handwriting-based Parkinson’s disease detection using image data. It employs hierarchical features by getting sequential architecture of convolutional layers and max-pooling layers. Pre-treated with resizing, rescaling and augmented random flipping of images to generalize. Robustness is provided with soft-max activation and Adam Optimizer for that model.