Vivien Chappelier commited on
Commit
5a429a0
1 Parent(s): cc7bb74

explain purpose of the method

Browse files
Files changed (1) hide show
  1. README.md +7 -1
README.md CHANGED
@@ -26,10 +26,16 @@ pipe = StableDiffusionXLPipeline.from_pretrained(model, vae=vae)
26
 
27
  For more information, please have a look at [the official demo](https://huggingface.co/spaces/imatag/stable-signature-bzh)
28
 
29
- ## Decoder Finetuning
30
 
31
  This model was finetuned from the original StableDiffusion-XL autoencoder with the additional objective of including a weak invisible watermark, following the procedure of [StableSignature](https://ai.meta.com/blog/stable-signature-watermarking-generative-ai/). This watermark is detectable by [IMATAG](https://www.imatag.com/)'s demo BZH decoder, available via API.
32
 
 
 
 
 
 
 
33
  ## Detecting the watermark
34
 
35
  For security reasons we cannot release the weights of the detector. Instead, you may check for the presence of a watermark with the [detect_api.py](https://huggingface.co/spaces/imatag/stable-signature-bzh/resolve/main/detect_api.py) script like this:
 
26
 
27
  For more information, please have a look at [the official demo](https://huggingface.co/spaces/imatag/stable-signature-bzh)
28
 
29
+ ## Purpose
30
 
31
  This model was finetuned from the original StableDiffusion-XL autoencoder with the additional objective of including a weak invisible watermark, following the procedure of [StableSignature](https://ai.meta.com/blog/stable-signature-watermarking-generative-ai/). This watermark is detectable by [IMATAG](https://www.imatag.com/)'s demo BZH decoder, available via API.
32
 
33
+ Compared to other watermarking solutions such as [invisible-watermark](https://github.com/ShieldMnt/invisible-watermark):
34
+ - the watermark is quite robust to many unintentional editing transforms such as cropping, resizing, changing contrast or brightness, and JPEG compression
35
+ - the watermark has zero overhead in terms of computational power
36
+ - the watermark is embedded in the model weights and not as a post-processing of the non-watermarked image
37
+ - the probability of detecting the watermark by chance on non-watermarked images may be computed analytically and limited to a specific threshold
38
+
39
  ## Detecting the watermark
40
 
41
  For security reasons we cannot release the weights of the detector. Instead, you may check for the presence of a watermark with the [detect_api.py](https://huggingface.co/spaces/imatag/stable-signature-bzh/resolve/main/detect_api.py) script like this: