ODE-based generative modeling: Learning from a single natural image

Jian Yue, Yan Gan, Lihua Zhou, Yu Zhao, Shuaifeng Li, Mao Ye

Abstract: Highlights•We proposed an ODE-based method for efficient, high-quality single-sample image generation.•We removed transitional noise between sampling stages, improving model efficiency.•We enabled customized image generation using a multimodal pre-trained model.•We achieved diverse, high-quality samples in extensive experimental evaluations.