Xiaoshan Yang
Xiaoshan Yang
Thank you for your reply. I tested the model trained with 3 stages. The evaluated results are still not good: Bleu_1: 0.127 Bleu_2: 0.032 Bleu_3: 0.007 Bleu_4: 0.002 Could you...
I actually only used the default hyper-parameter setting in the training and generation code. I will try to fix that. Thanks.
@niuzhiheng I have successfully tested feature extraction and fine tuning the CNN model with your caffe windows. Thanks for your help! When I rebuild the caffe on vs2012, I find...