search

Bartzi / see

Code for the AAAI 2018 publication "SEE: Towards Semi-Supervised End-to-End Scene Text Recognition"
GNU General Public License v3.0
574 stars 147 forks source link

where does the training params store? #12

Closed qnkhuat closed 6 years ago

qnkhuat commented 6 years ago

Hi, I wonder where new params store. and how can I use it for demo ?

Bartzi commented 6 years ago

trained params are stored in the *.npz files that are produced by the snapshotter (after each epoch, or after a given amount of iterations. This amount of iterations os configured with the --snapshot-interval command line switch). I also provided some of those files on bartzi.de.

The README also tells you how to use those models.

qnkhuat commented 6 years ago

oh. So it is the model which I specified in training command. In my case I use model_190000.npz?

Bartzi commented 6 years ago

If you want to to use the demo script on one of my models you should use model_190000.npz but if you want to use a model trained on your dataset, you will have to use one generated by the training. You can find it in the log folder that has been created for your train run.

Is that what you want to know? I'm not sure whether I get your question.

qnkhuat commented 6 years ago

yes. I want to use the model I trained

Bartzi commented 6 years ago

okay, you should find a model in the log folder. Look for any .npz files if there is nothing there yet, you'll have to wait or set the --snapshot-interval command line switch