Redundant output of the `box_predictor` and `class_predictor` of the second phase network about img2pose

[FunkyKoki]

By the way, these code are redundant as well, since the input params of this rcnn_loss are already positive proposals, labels and so on.

[vitoralbiero]

Hello @FunkyKoki,
Thank you for your interest in our work.
Yes, you could simplify the code to use just 1 class, but since we are using the standard Faster R-CNN implementation, the background class is required: num_classes (int): number of output classes (including background.

Hope this helps.

[FunkyKoki]

All right, thanks for your reply. 😃