Hi! Have you guys try out batch size larger than 1?I want to know how to implement it?

[moFang222]

When I try batch_size = 2
I encountered the following problem

So I checked the Dimension and the problem appears in mask_batch on line of code
rois_label, stu_feature, mask_batch = fasterRCNN(im_data, im_info, gt_boxes, num_boxes)

I don't know how to modify the code because it needs to involve some complicated changes.

[chenaifang]

Do you have implement it now? @moFang222

[twangnh]

@moFang222 @chenaifang Hi thanks for being interested in our research, we have not tried larger than one batch for faster r-cnn, but it is simple, you can implement it yourself.