Why does running the program using the data sets FLO,SUN,AWA2 not yield the results in the paper?

hanouticelina / lsa-zsl

Official PyTorch implementation for "Learning Semantic Ambiguities for Zero-Shot Learning".

Other

5 stars 1 forks source link

Why does running the program using the data sets FLO,SUN,AWA2 not yield the results in the paper? #1

Open ProgrammerBrcueWang opened 1 year ago

ProgrammerBrcueWang commented 1 year ago

Hello, author. When I ran your model, I could not get the experimental results in the paper on the data sets FLO,SUN and AWA2. Could you please provide the parameters when running FLO,SUN and AWA2?

hleborgne commented 1 year ago

All the parameters are in the Makefile. Using e.g make flo_trans you should obtain the results on FLO for transductive settings (up to possible small variation due to intrinsic randomness). Do you obtain significant differences ?