bobliu20 / faciallandmark_caffe Goto Github PK

batch_reader.py里的unit test可以通过，图像也可以存下来，但是放到net2的训练里就会出错，输入的图像size设的是40*40，"buffer2memory":False，不清楚什么原因

Hi,
This model can using face detection first right?

Hi,
This method the author use two-stage, the achievement just use the first stage?
Thank you.

Hi ，thanks for your gread source code
could you show me the train and validation format in txt.?

Hi,
Can you share me the net1 model? My email address is [email protected].
Thank you very much,

关键点欧几里得loss收敛时大概多少？训练有啥参数设置吗？我训练过一直不好收敛

i am not able to locate file "full_path_zf_bbox_pts.txt" in celebA dataset

In my caffe project do train command. Error： Unknown layer type：Python
can you tell me the reason?

I test the net2/_iter_150000.caffemode using caffe, but I can't get the correct result. The mistake may comes from anti-normalization. So, what's initial normalization method?

Where i need to make changes to detect more landmarks and training data is in .csv format do i need to modify it