Background: By considering current practices, problems, and prospects of image classification, emphasis is placed on investigating the medical leaf image dataset. In Ayurvedic medicine, the correct identification of medical plants is very important. However, this identification is very easily done by human beings. Botanists are trying to identify more species, which helps in recognizing the new species. Methods: The Swin-transformer model explained the impact on the medical leaf image dataset. The proposed approach of multi-head self-attention (MSA) mechanism is applied for the identification and classification of medical leaf images. Using hierarchical representation, at each stage resolution along with embedding dimension, features are extracted, which helps to visualize the classification of images. Results: The recommended model achieved accuracies of 98% using the Adam optimizer, the cross-entropy function and the softmax activation function. By augmenting the leaf image dataset, we escalate the dataset in order to enhance the accuracy and provide scalability. Conclusion: This study assessed the results showing that the Swin Transformer can enhance performance and has obtained good results in the dataset.