Update README.md
Browse files
README.md
CHANGED
@@ -22,7 +22,7 @@ This model demonstrates greatly improved performance over the umm-maybe detector
|
|
22 |
|
23 |
However, its performance may be lower for images generated using models other than SDXL. In particular, this model underperforms the original detector for images generated using older models (such as VQGAN+CLIP).
|
24 |
|
25 |
-
While the original umm-maybe AI art detector was trained on data scraped from Reddit, the data used for this fine-tune is either synthetic (generated by SDXL) and therefore non-copyrightable, or downloaded from Wikimedia and therefore meeting their definition of "free data" (see https://commons.wikimedia.org/wiki/Commons:Licensing for details).
|
26 |
|
27 |
# Model Trained Using AutoTrain
|
28 |
|
|
|
22 |
|
23 |
However, its performance may be lower for images generated using models other than SDXL. In particular, this model underperforms the original detector for images generated using older models (such as VQGAN+CLIP).
|
24 |
|
25 |
+
While the original umm-maybe AI art detector was trained on data scraped from image links in Reddit posts, the data used for this fine-tune is either synthetic (generated by SDXL) and therefore non-copyrightable, or downloaded from Wikimedia and therefore meeting their definition of "free data" (see https://commons.wikimedia.org/wiki/Commons:Licensing for details).
|
26 |
|
27 |
# Model Trained Using AutoTrain
|
28 |
|