comfyui-optical-flow

comfyui-optical-flow
★ 33

光流视频处理运动估计图像扭曲
提供计算并可视化相邻图像间光流的节点,并能将光流应用于同尺寸图像,实现帧间运动迁移与视频一致性处理。
💡 在视频帧序列中计算/可视化光流,或将运动场应用于同尺寸图像实现帧间迁移。
🍴 8 Forks💻 Python🔄 2024-05-22
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/b45acaa3411d
📄 README

About

This package contains three nodes to help you compute optical flow

between pairs of images, usually adjacent frames in a video, visualize

the flow, and apply the flow to another image of the same dimensions.

Most of the code is from Deforum, so

this is released under the same license (MIT).

Nodes

Compute optical flow

This node takes two images, prev and current, and computes the optical

flow between them using either the DIS (Dense Inverse Search) medium

or fine method, or Farneback. The images must have the same

dimensions.

Apply optical flow

This node takes an image and applies an optical flow to it, so that

the motion matches the original image. This can be used for example to

improve consistency between video frames in a vid2vid workflow, by

applying the motion between the previous input frame and the current

one to the previous output frame before using it as input to a sampler.

Visualize optical flow

This node takes an image and a flow and produces an image visualizing

the flow on top of the image. The image must be the same size as the

images used to compute the flow in the first place. It’s up to you

whether you use the “prev”, “current”, or an image you intend to apply

the flow to.