Input total amount of images in image limit, path to folder with images in load image batch, select model (I use only object detection models here), should work for segmentation too, enter N batch count and queue. For obb I generate masks outside and use in workflow 2.
Job iterator custom nodes allow to batch jobs without unloading models between each node which results in massive speed up. I use 50 it's fits images nicely in my ram without needing to cloes everything else.
Workflow embedded in the image so you can drag-and-drop image onto ComfyUI canvas.
Description
FAQ
Comments (4)
where to get suirano.pt ?
I uploaded yolo models that i trained, you can see it in suggested resources.
Your workflow worked better than my attempts using FaceDetailer, in particular I used the 2nd image workflow, just reducing the threshold. Only thing of note is that the detector returned a mask in the wrong position, so I used this detector instead.
If you need general watermark detection this model is quite good: https://huggingface.co/spaces/fancyfeast/joycaption-watermark-detection/tree/main
File far5y1y5-8000.pt
Found it in article https://civitai.com/articles/10464/wacky-misadventures-in-detecting-watermarks
I moved away from this workflow because i dislike comfy and wrote app for it. It's ugly but suits me better.