better i think, but still.... faces are too similar xD
better i think, but still.... faces are too similar xD
Fbut i approve! 
waiting for the video
waiting for the video
Fwell these guys are similar :d
Fuploading tutorial
Fvideo almost ready. waiting youtube to check copyright
FFlmao i was thinking you needed to train a face on SD or something, not just a single photo!
Fhow many vram is needed?
Fonly 3?
Fwhere did you take the model from?
Fthe inswapper
Fit could be malicious!
Flet me find original source. on windows i couldnt make it work with gpu
Fok onnx is safe
FAnother mitigation approach consists in limiting what models are allowed to do, to contain and disrupt potential malware:
Some model formats are more restrictive than others, like ONNX or TFlite for instance (when avoiding custom operators).
Some model formats are more restrictive than others, like ONNX or TFlite for instance (when avoiding custom operators).
FGitHub
a CLI that provides a generic automation layer for assessing the security of ML models - Azure/counterfit
FPyTorch Forums
I am currently organizing a NeuroIPS competition, in which participants might be submitting pytorch models to our evaluation server for the leaderboard. Is there a secure way of serializing/loading untrusted pytorch models? Are there alternatively to pickle, which can be insecure? Is there a pytorch model format we can insist upon that is secu...
Fye onnx is safe
Fso it's like a safetensors
Fyes
Fbut wait
FHiddenLayer | Security for AI
In this article, we'll show you how ransomware can be automatically launched from a machine-learning model.
Flook the table
Factually safetensors is safe, onnx is not
Finteresting
Fother places tells it is safe
Fwell this article looks very technical xD
FFGitHub
swapper = insightface.model_zoo.get_model('inswapper_128.onnx', download=True, download_zip=True) cannot download the model from given url. RuntimeError: Failed downloading url https://gith...
Fyup, i've seen those
Fthe problem is... this model is now private, and probably you got it from a link where you can't assure if its safe or not?
Fwe would need the hash of the original model to compare
Fye that was the thing i was going to do
F
Fwell at worst you can run on colab
Fthat's true
Fi got kaspersky internet security
Fdidnt give any warning or such
Fcan we use another model from here? https://github.com/deepinsight/insightface/tree/master/model_zoo
Fi think it is possible
Fand they are official downloads
Fwell nevermind, models work with insightface library, but "roop" repo needs the inswapper
Fbecause it's more than just insightface
Flooks like buffalo_l is the model for face analysis and inswapper_128 is for swapping
FFthis is another swapper model
Fthis is a colab in case you want to try: https://colab.research.google.com/github/neuralchen/SimSwap/blob/main/SimSwap%20colab.ipynb
FFlooks easy too
Fyou can even train the model yourself, with your own face
Fnice find