SinFusion: Training Diffusion Models on a Single Image or Video
This addresses the need for efficient, single-input manipulation in image and video processing, offering a novel approach beyond large-scale training.
The authors tackled the problem of adapting diffusion models to manipulate a single image or video, which typically require large datasets, by training a model on a single input. The result is SinFusion, which can generate diverse new video samples, extrapolate videos in time, and perform video upsampling, tasks not achievable by current methods.
Diffusion models exhibited tremendous progress in image and video generation, exceeding GANs in quality and diversity. However, they are usually trained on very large datasets and are not naturally adapted to manipulate a given input image or video. In this paper we show how this can be resolved by training a diffusion model on a single input image or video. Our image/video-specific diffusion model (SinFusion) learns the appearance and dynamics of the single image or video, while utilizing the conditioning capabilities of diffusion models. It can solve a wide array of image/video-specific manipulation tasks. In particular, our model can learn from few frames the motion and dynamics of a single input video. It can then generate diverse new video samples of the same dynamic scene, extrapolate short videos into long ones (both forward and backward in time) and perform video upsampling. Most of these tasks are not realizable by current video-specific generation methods.