search

tengshaofeng / ResidualAttentionNetwork-pytorch

a pytorch code about Residual Attention Network. This code is based on two projects from
667 stars 165 forks source link

About multi-label #6

Closed josianerodrigues closed 6 years ago

josianerodrigues commented 6 years ago

Hi @tengshaofeng, Do you know if this model can process multi-label datasets, like NUSWIDE? Any idea how to do it? Thank you.

tengshaofeng commented 6 years ago

multi-label means a image have several labels? if yes, than you can modify the softmax in last layer to sigmoid, each neuron is connected with sigmoid, and each neuron is responsible for one label.

josianerodrigues commented 6 years ago

Yes, multi-label means a image with more than one label. In that case, do you suggest any specific pytorch loss function?

tengshaofeng commented 6 years ago

nn.BCELoss