BriefGPT.xyz
Nov, 2024
扩散模型在感知任务中的规模属性
Scaling Properties of Diffusion Models for Perceptual Tasks
HTML
PDF
Rahul Ravishankar, Zeeshan Patel, Jathushan Rajasegaran, Jitendra Malik
TL;DR
本研究解决了扩散模型在视觉感知任务中的应用问题,通过将深度估计、光流和模态分割等任务统一在图像到图像转换的框架下,提出了一种新的训练和推理方案,以优化计算效率。结果表明,这些模型在使用显著更少的数据和计算量时,能达到与最先进方法相当的竞争性能。
Abstract
In this paper, we argue that iterative computation with
Diffusion Models
offers a powerful paradigm for not only generation but also
Visual Perception
tasks. We unify tasks such as depth estimation, optical flow,
→