yehao
yehao
You can use (http://ethereon.github.io/netscope/#/editor) to view peleenet-ssd network structure, you will find stage4_tb/ext/pm2/res layer is used twice to generate ext/pm1_mbox_loc layer and ext/pm2_mbox_loc layer(conf layer/priorbox layer is also the case)....
mobile-net ssd also follows this design pattern, I will update later.
No, batch size can affect training time and has no direct relation with model performance.
You can look [it](https://www.zhihu.com/question/32673260)
@foralliance model performance 依赖于模型本身，batch size只是一个超参而已，如果你修改了batch size，再选定其它合适的超参比如base_lr，它最终的效果是一样的，一般而言，使用一个大batch只是训练的更快，更容易出结果，并不会从根本上决定模型的性能
@foralliance 我看了这个回答，默认accum_batch_size固定，对于无BN层/BN层参数固定情况下，batch size不影响模型性能这个观点我是认同的
You download a wrong link. I guess you download pelenet_inet_acc7243.caffemodel, It is a classification model rather than detection model.
@lucheng07082221 you can test it by yourself.
what is c++ implementation? Do you refer to peleenet network or framework(like caffe or ncnn)?
@Robert-JunWang Hi, I have two problems: - I found that you have uploaded a merged model, what is the principle of this? merged BN with convolution layer change original convolution...