search

GriffinLiang / vrd-dsr

Code for Visual Relationship Detection with Deep Structural Ranking (AAAI2018)
121 stars 32 forks source link

Changing the batch size #14

Open sukrutrao opened 6 years ago

sukrutrao commented 6 years ago

Is it possible to change the batch size while training? I see that there is a _batch_size variable in the VRDDataLayer class, but the only place it is used seems to be here for setting the number of epochs. While actually doing the forward pass, it seems that only one image is used.

Is this the case? Will training code have to be modified to allow a higher batch size?

Thank you.

GriffinLiang commented 6 years ago

I think it's possible to use more data in one batch. You need to modify the codes related to roi pooling and the ranking loss.