TMPVT: few-shot classification of Tibetan medicinal plants using sision transformer

In this paper, a novel Transformer Vision framework for Tibetan medicinal plants (TMPVT) is studied, aware of a few-shot Tibetan medicinal plants’ (TMP) image classifications and recognition. Tibetan medicinal herbs are valuable natural medicinal plants that have important healthcare values and are useful in treating a wide range of diseases. In the system we studied, we utilized the self-attention mechanism in the Vision Transformer model to identify and classify Tibetan medicinal herbs. This approach allowed us to achieve accurate classification for image annotation of these herbs, including the localization of the herbs’ targets and their respective categories. In addition, we constructed a dataset that contains annotated images of various types of Tibetan medicinal herbs extracted from a database of real plant images. We simulated and evaluated the proposed solution and demonstrated that the performance of TMPVT for solving the problem of classifying images of Tibetan medicinal herbs with few samples is significantly better than other available options.