Changan Wang comments

Results 15 comments of


                                            Changan Wang

error in 'make'

@czy36mengfei cmake的输出看一下

error in 'make'

@czy36mengfei 看一下python命令默认是哪个环境？tensorflow，TF_LIB都没找到

why no weight decay used ?

same question. have anyone tried weight_decay?

I have one issue, when i train model. Key conv1/7x7_s2/bn/beta/Momentum not found in checkpoint is not found.

@chanyixialex Momentum is not a part of model param, it's used by optimizer, so isn't included in the provided checkpoint

train SSD on my own data,but the rusult of detector is so bad

Maybe you can try [this](https://github.com/HiKapok/SSD.TensorFlow), it's easier to train.

train SSD on my own data,but the rusult of detector is so bad

@LJXLJXLJX thank you

training from imagenet can't converge

me too! have you found a solution?

training from imagenet can't converge

@ylqi007 [this works](https://github.com/HiKapok/SSD.TensorFlow/)

The generated graph is degenerate

Hi, here is also one Tensorflow SE-Net with pretrained wieghts converted from Caffe. See [here](https://github.com/HiKapok/TF_Se_ResNe_t) if you are interested in.

where is xception like network code which is written in original paper?

Hi, I have also tried the original Xception backbone in my re-implementation and found it worked too. And the forward time in training for backbone network is about 20ms. But...