Images play a crucial role across diverse fields; however, traditional image enhancement methods possess certain limitations. In contrast, the Diffusion Model presents distinct advantages in image generation. IADL (Image Augmentation Driven by Language) utilizes large vision and language models to generate natural language descriptions and augment training data. The research focuses on the field of botany, especially the identification of precious flowers. Experiments were conducted on the 102 Category Flower Dataset, and the performance of IADL was compared with traditional image enhancement techniques. The results show that traditional techniques still perform better, but Img2Img outperforms Pix2Pix, highlighting the efficacy of feature extraction networks. The IADL method achieves notable results as a generative approach for image creation.