This project was submitted to AI for Good Global Summit.
Recreating Photos Portraying Severe Effects of Climate Change via CycleGAN
The importance of climate change is difficult to neglect regarding current societal status. Numerous typhoons and hurricanes have swept densely populated areas already, and the state of California is undergoing a severe wildfire crisis. In fact, a recent report from the Intergovernmental Panel on Climate Change has determined that drastic and rapid changes are required in order to avoid increasing climate change risks for nature and human systems. To address these issues and highlight the severity of the effects of climate change, inspired by the novel research by V. Schmidt et al., a CycleGAN model is introduced to generate a fake image of flooded and collapsed houses. By training the model on images of flooded houses and collapsed houses due to earthquakes, the model can generate effects on new house images. The eventual goal of the project is to enable future generations to see and refer to the model in the work of restoration and be alerted. If given enough resources, the future work is to evaluate more accurate performance by providing more images as input and fine-tuning the model once again.
View Project Paper