The goal of this point is to find a face of any type of generation, including realistic, anime, and anthro characters.
PLEASE NOTE THE DATABASE TYPE DOES NOT MATTER. I have adjusted each version to include a different database version for search ability, but these are diffusion agnostic. It's all about detecting a face no matter what.
This is a bbox detection YOLOv8 model. This does not improve your image on its own and must be used as a part of the 'adetailer' extension for WebUI, found at https://github.com/Bing-su/adetailer or 'uddetailer' found at https://github.com/wkpark/uddetailer. It is to be used to find any type of face in a image so that you can infill that part of the image. Right now it is better suited for finding faces that the other pre-installed databases seem to miss. (ex. face_yolov8n.pt, face_yolov8s.pt, mediapipe_face_full). If those did not work for you, mine probably will.
This is my first model and first item I have made I feel confident in that others may find it useful. I found out that these adetailer databases are actually part of something that I was already working on.
Right now it is not trained on realistic animals or common mistakes, but I have seen my desired success rate already. Please let me know if you continually get a false positive (what you and I know probably should not be a face).
I realized to late that I focused to hard on if I could, that I didn't stop to think if I should. Now I am making a new data set and I have to re-tag all my 'facial' data... The next one will have classes. There are some rumblings that adetailer may be able to pick and choose which classes of detection to remake.
Later if I can work this out I'll probably move on to segmentation, time willing. I have met challenges with this with my tools.
Please feel free to provide your feedback on which faces it seems to miss or what false positive you find, as I want to continue to improve this model to find Only faces :)
Description
V1.1 differs from v1.0 as I used better images for tagging realistic humans, and removed symbiotes from my data set, as they seemed to throw off the model and not even find symbiotes reliably. Strangely enough, this seems to find them better.
FAQ
Comments (9)
finally something that can detect non-human faces
awesome, how to install it?
Oh my gosh I hope you love this plugin.
From the github: https://github.com/Bing-su/adetailer :
--Put your ultralytics yolo model in webui/models/adetailer.
This model is a bbox detection model, I hope some day I can make a segmentation model, but that takes a lot longer to tag images for segmentation than it is to draw the box over the item and label it.
@l0ud_ninja Haven't searched enough to be 100% sure but you can train segmentation models with yolo too right? Haven't seen a seg model for faces and genitals yet so that's something that could be useful.
Thank you for working on this!
@Jilhbril 100% agree. adetailer can handle both detection models like mine and segmentation models. After I get this down a little better on detection I was going to move over to segmentation of the same model(s).
ffjggrtbjibv made one that has worked for me (https://civitai.com/models/132388/penis-pussy-adetailer-model), However I accidentally made the leap from horrible nsfw to something I'm satisfied with, and then their model didn't help me out much.
If you know of a repository of images that are procedural generated with lots of mistakes like faces... or ... other things....... let me know :)
so.....this is a Adetailer model right ?
But there is already a model like that mate ..
it's called "face finder" and works pretty well...
Why this model so ? you're not very precise on your informations.... Is it for Adetailer ? is it a model ? it is a part of another face tracker ? be more specific when you post something like that bro...Thanks :)
Thank you for your feedback. I'm not 100% on your confusion but anything helps as this is my first posting.
This is exclusively for the adetailer plugin for WebUI, it will not function to help with image generation on it's own.
I am not aware of another face finder that works on every type of face, inside or outside of adetailer. Of the adetailer models I have found, most either focus on realistic humans or anime humans. Of those I have seen the data sets for anime also trend towards females or youth. I wasn't to satisfied with the results I found there, as sometimes I needed improvement on more male faces, or just animals that were to eldritch, possibly due to poor prompting on my part, for the finder to detect.
This was my attempt to make an adetailer module that can help finding... any type of face. Right now the data set nicely does better with more anthro or animal faces, so it would be suitable to use the default face finders for other faces anyhow. Maybe I'll focus it on different face types. Not sure, just was excited because it overlapped with what I was already interested in making: a detection and classification model.
I've been on the lookout for more face detection models, but do you think going for every face might be a bit too ambitious? Would it make more sense to train separate models for each face type you're trying to detect? Maybe one for realistic human, one for realistic animal, one for anime/cartoon human, and one for anime/cartoon animal/furry?
Regardless, I'll give this a shot to see how it works. Thanks for making it.
I was so focused on if I could, that I didn't focus on if I should. Thank you for the re-enforcement.
I am actually going to separate out my datasets for this exact reason. I have a robust dataset partially already, but it would be better if I had lots of images of bad, procedural generated faces.... where to get those...