These days, pharmaceutical companies are big fans of medicinal plants. Why? Because they are cheap and have fewer side effects than regular drugs. But, identifying these plants can be tough, time-consuming, and prone to errors. That's why researchers are excited about automatic classification of therapeutic plants. One cool method uses images of leaves and plants to do this.

This study used a big dataset of medicinal plants and their leaves. The first step was to make the images clearer with a technique called contrast-limited adaptive histogram equalization (CLAHE). Then, they used a special type of computer network called Attention-based Enhanced Local and Global Features Network (AELGNet) to look at these images. This network extracts and classifies important features.

AELGNet has two main parts. First, it uses three MBConv modules to find base features in the images. These features are then split into four non-overlapping patches for local feature extraction. At the same time, AELGNet also looks at base features for global feature extraction. To make these features pop out even more, the network uses residual channel-wise and spatial attention on the patches and global features.

The researchers tested AELGNet using a dataset of Indian medicinal plants. The results were impressive: AELGNet was 99. 71% accurate, with 99. 80% precision, 99. 75% recall, and 99. 77% F1 score. This means it was better than 14 other methods, improving accuracy by 2%-10%.

This study shows that AELGNet is a powerful tool for quickly and accurately identifying medicinal plants and leaves. It can be used in medical and industrial settings.

Plant-Spotter: A Smart Way to ID Medicinal Plants

Actions