Transframer: Arbitrary Frame Prediction with Generative Models
It addresses the challenge of multi-task computer vision for researchers and practitioners by proposing a unified framework, though it is incremental as it builds on existing U-Net and Transformer components.
The paper tackles the problem of unifying diverse vision tasks through probabilistic frame prediction, resulting in a generalist model (Transframer) that achieves state-of-the-art on video generation benchmarks, competitive performance on few-shot view synthesis, and promising results on 8 tasks without task-specific components.
We present a general-purpose framework for image modelling and vision tasks based on probabilistic frame prediction. Our approach unifies a broad range of tasks, from image segmentation, to novel view synthesis and video interpolation. We pair this framework with an architecture we term Transframer, which uses U-Net and Transformer components to condition on annotated context frames, and outputs sequences of sparse, compressed image features. Transframer is the state-of-the-art on a variety of video generation benchmarks, is competitive with the strongest models on few-shot view synthesis, and can generate coherent 30 second videos from a single image without any explicit geometric information. A single generalist Transframer simultaneously produces promising results on 8 tasks, including semantic segmentation, image classification and optical flow prediction with no task-specific architectural components, demonstrating that multi-task computer vision can be tackled using probabilistic image models. Our approach can in principle be applied to a wide range of applications that require learning the conditional structure of annotated image-formatted data.