Worldwide, the number of reported instances of Parkinson's disease (PD) has been on the increase. A lot of medical practitioners benefit from early diagnosis thanks to computer-aided solutions. The current research study demonstrates that computer-assisted diagnostics makes extensive use of deep learning technologies especially using transfer learning approaches. These techniques make extensive use of baseline models like VGG16, Inception, ResNet, MobileNet, and many more. However, in light of the present state of research, a comparative analysis of popular baseline models on the publicly available PD dataset is lacking. Therefore, we are driven to conduct a comparative study of 9 most common baseline models for PD diagnosis utilizing transfer learning. This paper also conducts an extensive results analysis of baseline models with or without finetuning. The experimental analysis shows that the VGG16, ReNet, Inception, DenseNet shows very promising results on the publicly available dataset.