about fine tuning decision network

ShuaiLYU / Deep-Learning-Approach-for-Surface-Defect-Detection

A Tensorflow implementation of "Segmentation-Based Deep-Learning Approach for Surface-Defect Detection"

734 stars 258 forks source link

Open roywang2011 opened 5 years ago

roywang2011 commented 5 years ago

When training decision net after training segment net, did the author add the operation of freezing segment convolution layers?I didn't find it.

ShuaiLYU commented 5 years ago

Different training op corresponds to different params. you can find it in the 'build_model' function in the model.py.

roywang2011 commented 5 years ago

Thank you very much!

JiangDing1995 commented 5 years ago

@roywang2011 Hi,friend.I didn't find the freezing layers either.Does that mean we don't have to modify the code?

roywang2011 commented 5 years ago

There is a param in the 'build_model' function.You may need to find it more carefully.