Model does not converge on larger dataset

MaybeShewill-CV / CRNN_Tensorflow

Convolutional Recurrent Neural Networks(CRNN) for Scene Text Recognition

MIT License

1.03k stars 388 forks source link

Model does not converge on larger dataset #367

Closed guddulrk closed 5 years ago

guddulrk commented 5 years ago

Hi, I am trying to train the network on my own data-set, which has 0.5 million images of synthetic text words. I have increased the depth of the network (added more layers), changed learning rates, optimizers and kernel sizes. I also removed the max-pooling layer and used strided convolutional layers. But the network is not converging. Can you help me, what should I change so that the network can converge? The training cost does not reduce than the 12.

Thanks.

MaybeShewill-CV commented 5 years ago

@guddulrk I am afraid I can do nothing to help you with this. Remote assistance for such kind of a problem will not work in my opinion. Sorry for this:)

guddulrk commented 5 years ago

No problem. Thanks