How to batch save openpose images??
This is driving me nuts. I'm using this: https://toyxyz.gumroad.com/l/jydvk
Which is excellent, basically I'm trying to generate the pose images of a person running, and then use those as the input for ControlNet image sequence 2 image sequence script v2.
But damn if I can't figure out how to output the poses as numbered images. The wierd and annoying thing is I managed to do it last night by accident and had to copy out all the images. But I can't do it again!
Does anyone have any technique for generating just the pose images from a collection of frames? Obviously not one at a time.
Thanks!
No, and even worse, I know it's possible somehow but I have tried every variation on the settings that I can think of.
I may end up just coding something up to do it myself, because it seems like it would be helpful to the community.
Last time it was something like doing a normal img2img batch with controlnet turned on, but for openpose I got black detectmaps only, and the others were deepfried. Maybe it got fixed since in an update, and I saw controlnet itself has a batch option now, so I'll have to check.
Does anyone have any technique for generating just the pose images from a collection of frames? Obviously not one at a time.
I used https://github.com/CMU-Perceptual-Computing-Lab/openpose/releases (Openpose) to generate images (contain keypoint) sequence from a video.
I run this in dev console of files list of model I would to download (go https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main, press F12 or whatever to go dev console, paste following)copy(`wget -c ${Array.from(document.querySelectorAll('a[download]')).map(a=>a.href).filter(a=>!a.match(/.gita/)).join(' ')}`)
now you have wget -c command contains all files in page (warning this is very basic and not expect spaces in file names :) )
sorry missed tread :(