生成式数字作物助力农学类课程教学新方法

A Novel Method of Utilizing Generative Digital Crops to Aid in Teaching Agricultural Courses

摘要: 农学类课程的现有讲授式课堂教学缺乏直观感受，而实验室或现场教学对条件要求高，费时费力。针对上述问题，文章提出通过生成式数字作物营造低成本且具有现场沉浸感的教学环境，从而助力农学类课程教学的新方法。该方法采用Transformer和扩散模型(Diffusion Model)这两种先进的深度学习架构，用于模拟作物生长的各个阶段，并生成对应高质量作物图像，模拟作物从种子到成熟的过程，通过不同分镜头实现作物生长过程的控制，能从宏观和微观角度观察作物形状、生理结构、生长机制。通过构建水稻、辣椒、葡萄三种典型示范数字作物过程，证明了这种教学方式的高效性。

Abstract: The current pedagogical approach in agricultural education, characterized by traditional lecture-style classroom teaching, often lacks an intuitive experiential component. Conversely, laboratory or on-site instruction necessitates high resource conditions and tends to be both time-consuming and labor-intensive. To address these challenges, this paper proposes a novel method that employs generative digital crops to establish a low-cost and immersive teaching environment, thereby enhancing the educational experience in agricultural courses. This innovative approach leverages two advanced deep learning architectures: the transformer model and the diffusion model. These technologies are utilized to simulate various stages of crop growth and generate corresponding high-quality images of crops. The method effectively replicates the developmental process from seeds to maturity while allowing for control over different aspects of crop growth through diverse perspectives. This enables observation of the shape, physiological structure, and growth mechanisms of crops from both macro and micro viewpoints. The efficacy of this teaching methodology has been validated through the construction of three representative demonstration processes involving digital crops: rice, chili peppers, and grapes.