my_val loss is high with your configure setting for your generate plates

qjadud1994 / CRNN-Keras

CRNN (CNN+RNN) for OCR using Keras / License Plate Recognition

MIT License

527 stars 191 forks source link

my_val loss is high with your configure setting for your generate plates #19

Open ghost opened 5 years ago

ghost commented 5 years ago

Hi @qjadud1994 I trained your Plates with your setting training but the loss is't reach that you have achieve 0.01 , but my val loss was 26 , why ?

OS : Ubuntu 16.04 LTS GTX : 1080 ti RAM :32 GB

qianjinhao commented 5 years ago

I have the same problem and don't know why

ghost commented 5 years ago

I tested this problem, I've solved this problem with number of large dataset. around 500k images got the val_loss 0.12.

allanruin commented 5 years ago

@mzeynali hi, how long does it take you to get val_loss to 0.12, I try a dataset which contains around 700k images, and use Nvidia M40 , at least the first 9 hours, model's val_loss stay around 17. I try to load the checkpoint model weights and see the prediction, it is garbage, something like '.8****', and all the prediction is the same value. Have you encounter this problem? Could you share some tips, thank you!