Retraining for line with more than two words

MaybeShewill-CV / CRNN_Tensorflow

Convolutional Recurrent Neural Networks(CRNN) for Scene Text Recognition

MIT License

1.03k stars 388 forks source link

Closed eravallirao closed 5 years ago

eravallirao commented 5 years ago

Can this architecture be used for longer text, If so what changes are needed to be made in order to get better accuracies.

MaybeShewill-CV commented 5 years ago

@eravallirao 1.Model input tensor size and final sequence length:)

eravallirao commented 5 years ago

Are there any optimal number which you can suggest for?

MaybeShewill-CV commented 5 years ago

@eravallirao That depends on your data:)

eravallirao commented 5 years ago

b97d950fe186a99ad4c9869afeb68183 My data is of this type, can you please suggest something

eravallirao commented 5 years ago

Do I need to change the LSTM layer from 256 to 512 and increase the layers of is the present one good enough

MaybeShewill-CV commented 5 years ago

@eravallirao No need to change that in my opinion:)