[Question] EfficientAD is patented by MVTec in Japan #2526
-
It seems that EfficientAD is patented by MVTec, at least in Japan. https://patentscope2.wipo.int/search/en/detail.jsf?docId=JP394418688&_fid=EP393389056 |
Beta Was this translation helpful? Give feedback.
Replies: 8 comments 1 reply
-
Are you sure if it is Efficient AD? I cannot see the details due to the language barrier. However, when I look at the list of the authors of the patent documents, I can see they don't match the that of Efficient AD authors. Paul Bergman, for example, is the author of the dataset, some previous mvtec papers, but he is currently not with mvtec anymore. |
Beta Was this translation helpful? Give feedback.
-
Also, the EfficientAD in Anomalib was implemented by @alexriedel1, by following the details in the paper, which is publicly available on ArXiv. |
Beta Was this translation helpful? Give feedback.
-
@samet-akcay |
Beta Was this translation helpful? Give feedback.
-
It is also patented in europe under this patent. It didnt fully read it yet and dont know if it would affect EfficientAD |
Beta Was this translation helpful? Give feedback.
-
Thanks for sharing @alexriedel1. I could contact the authors to see |
Beta Was this translation helpful? Give feedback.
-
Quick update on the issue. I had contact with the authors of the paper. Comment by MVTec Software GmbH: So for now I would not recommend to further make use of this architecture unless used for small private projects |
Beta Was this translation helpful? Give feedback.
-
Our plan is to implement GLASS and GLAD papers into Anomalib, both of which already outperforms EfficientAd. Once added, these models could be used instead of EfficientAd. |
Beta Was this translation helpful? Give feedback.
-
That some lovely news. |
Beta Was this translation helpful? Give feedback.
Quick update on the issue. I had contact with the authors of the paper.
Comment by MVTec Software GmbH:
In general, GCAD as well as EfficientAD can fall within the scope of the patent. Consequently, we strongly discourage the use of open-source implementations of the algorithms, e.g. provided within the anomalib library, to avoid IP conflicts and legal consequences.
So for now I would not recommend to further make use of this architecture unless used for small private projects