Cadene / vqa.pytorch

Visual Question Answering in Pytorch
716 stars 177 forks source link

Can we directly train the model on VisualGenome? #28

Open ghost opened 6 years ago

Cadene commented 6 years ago

No, you have to modify the visual genome dataset to be split into at least a train and a val splt