Hello,
Thank you for releasing DDAMFN and the accompanying paper — the accuracy/parameter trade-off is
remarkable.
We are evaluating facial expression recognition models for an academic project that may later
have a commercial component, and we would like to include DDAMFN in our comparison. However, we
could not find a LICENSE file in the repository. Without an explicit license, the default legal
position is "all rights reserved", which prevents us from using the code or weights beyond
reading them.
Would you consider adding a license file (for example MIT or Apache-2.0, as POSTER V2 and DAN
use)? Alternatively, could you clarify under what terms the code and pretrained weights may be
used?
We would also appreciate knowing whether the released weights depend on a backbone pretrained on
MS-Celeb-1M, since we are documenting the data provenance of every model we evaluate.
Thank you for your time.
