INR Smooth:在扩散模型上基于框架间噪声关系的平滑视频合成
Cuihong Yu1, Cheng Han1, Chao Zhang1
1School of Computer Science and Technology, Changchun University of Science and Technology, Changchun, China.
本研究介绍了INR Smooth,这是一种新的视频平滑策略,通过解决框架间的不一致性来增强文本到视频 (T2V) 生成. 该方法可以提高时间一致性和文本对齐,而无需额外的计算资源.
科学领域:
- 人工智能的人工智能
- 计算机视觉 计算机视觉
- 机器学习 机器学习
背景情况:
- 文本到视频 (T2V) 生产面临着像不一致和文本不整齐等挑战,阻碍了视频的流性.
- 现有的光滑方法往往会牺牲背景纹理和艺术表达.
研究的目的:
- 提出INR Smooth,一个视频平滑策略,解决用于改进T2V生成的框架间噪声关系.
- 开发基于培训和无培训的视频平滑编辑方法.
主要方法:
- 基于框架间噪声关系的平滑策略.
- 基于培训的方法:同时限制噪声和平滑损失功能.
- 没有培训的方法:DDIM反转用于增强文本对齐.
主要成果:
- 在文本对齐和时间一致性方面有显著的改进.
- 在平滑的过渡和艺术风格的描绘中表现出色.
- 无训练和零射击微调方法不需要额外的计算资源.
结论:
- 在T2V任务中,INR Smooth有效地增强了视频光滑.
- 提出的方法既提高了视觉质量,也提高了对文本提示的坚持.
- 可访问的实现,提供源代码和演示.
更多相关视频
12:15Image Processing Protocol for the Analysis of the Diffusion and Cluster Size of Membrane Receptors by Fluorescence Microscopy
Published on: April 9, 2019
06:55Synthesis of Cyclic Polymers and Characterization of Their Diffusive Motion in the Melt State at the Single Molecule Level
Published on: September 26, 2016
相关概念视频
Reconstruction of Signal using Interpolation
Frames: Problem Solving II
Uniform Depth Channel Flow
Uniform Depth Channel Flow: Problem Solving
Accelerating Fluids
The motion of the liquid within this infinitesimal cylinder is considered to obtain the pressure difference. Three vertical forces act on this liquid:
Boundary Conditions: Lossless Lines
At the receiving end, the boundary condition states that the voltage equals the product of the receiving-end impedance and current. This relationship is expressed as a function of the incident and...
