Rendering Extreme Weather Phenomena Using Generative AI Base on ControlNet

Generative artificial intelligence (AI) has rapidly advanced in recent years, achieving significant breakthroughs in computer vision and image processing. However, there are still many challenges in rendering complex scenes, particularly generating and controlling images for extreme weather conditions. In this work, we propose a novel generative model based on ControlNet and achieve precise control and high-quality rendering of extreme weather features by combining a sparsely connected Squeeze-and-Excitation module with a zero-convolution design. The proposed model excels in generating highly realistic extreme weather scenes, including heavy rain and snow, while significantly enhancing the ability to fine-tune weather characteristic parameters. Experimental results validate the model's exceptional generation quality and robust control stability, which not only realistically reproduces complex weather scenes but also allows flexible adjustments to weather characteristics such as rainfall intensity, snowflake density, and flame range. Compared to previous methods, the model demonstrates substantial improvements in image detail fidelity, weather feature consistency and generation efficiency. This technology shows immense potential for applications in disaster simulation and early warning systems.