Hi, I face the same problem now. It seems to me that there is an issue in sending the batch into forward_backward and further processing of it in the code from the tutorial. The shape [1, 128] is just one array from the entire batch of [8, 128] which is based on your batch_size = 8 . For instance, try batch_size = 4 and you will see [4, 128] instead. The training code for Faster RCNN is made in a different way to the one for SSD (that runs correctly). I am now attempting to get rid of that, but it is not ready yet.
Though, it would be nice to hear the comments from MXNet team... --- [Visit Topic](https://discuss.mxnet.apache.org/t/object-detection-finetune-f-rcnn-models/6580/2) or reply to this email to respond. You are receiving this because you enabled mailing list mode. To unsubscribe from these emails, [click here](https://discuss.mxnet.apache.org/email/unsubscribe/8d7b2b4842bf66ae788623113fd289aaa48f1cb8b4608324d44bec1774e00c98).
