Meta,也就是之前的Facebook,去年4月推出的Segment Anything模型非常好用,我在多个ComfyUI工作流中利用它实现了精准的"抠图"。时隔一年多,升级版Segment Anything 2来啦。它不仅在图像分割速度和效果上有显著提升(性能提高了近三倍),更是从图像领域扩展到了视频领域,能通过存储和利用之前处理过的视频帧信息,为动态视频内容提供实时、精准的对象分割。
在这个视频里,我将从模型和插件的下载安装开始,一步步带你完成单张图片内物品的标记、视频内的标记、以及视频内多个物品同时标记的工作流!
插件:https://github.com/kijai/ComfyUI-segm...
https://github.com/kijai/ComfyUI-KJNodes
模型:https://huggingface.co/Kijai/sam2-saf...
工作流:https://drive.google.com/drive/folder...
Watch video ComfyUI+SegmentAnything2:不惧遮挡、支持多目标,实现视频抠图自由!附详细的下载安装和工作流搭建流程 online without registration, duration hours minute second in high quality. This video was added by user SweetValberry 01 August 2024, don't forget to share it with your friends and acquaintances, it has been viewed on our site 2,41 once and liked it 7 people.