Hi,
the only place where I find an implementation of DenserNet is in Feature_branch.py. Anyway I am not able to understand how is the One-Stage Feature Decoder mentioned in paper implemented here; as I only see a fc layer on top of the feature branches that seems to be trained for classification and not for retrieval,
Can you give some indications in merit?
Hi, the only place where I find an implementation of DenserNet is in Feature_branch.py. Anyway I am not able to understand how is the One-Stage Feature Decoder mentioned in paper implemented here; as I only see a fc layer on top of the feature branches that seems to be trained for classification and not for retrieval, Can you give some indications in merit?