sercant/mobile-segmentation

Training my own data

hanseun opened this issue · 2 comments

Hi Author!

I'm trying to train using my own data.
I made tfdata and tried to excute train.py.
But I can't define "tf_initial_checkpoint".
Could you please help me out?

Hello @hanseun,

Can you add the error that you are getting while trying to define tf_initial_checkpoint?

The issue might be one of the following:

  1. If your class count is different, you haven't set initialize_last_layer to False.
  2. The checkpoint you are trying to restore is using DPC and you haven't specified that you want to use DPC by setting dense_prediction_cell_json parameter or vice versa.

If you send me the error message I can help you better.

thanks!