search

githubharald / CTCDecoder

Connectionist Temporal Classification (CTC) decoding algorithms: best path, beam search, lexicon search, prefix search, and token passing. Implemented in Python.
https://towardsdatascience.com/3797e43a86c
MIT License
817 stars 182 forks source link

best_path is better than beam_search in my experiments，is there something wrong with my network？ #14

Closed shaoming20798 closed 4 years ago

shaoming20798 commented 4 years ago

Hi, I use cnn + lstm + ctc in my license plate recognition system。And I use your python scripts to test the precision of "best path decoding" and "beam search decoding"。The result is "best path" got 92.3% while "beam search" got 91.7%。In my experiments, time step equals to 21，and the total num of labels equals to 77。 Could you give me some advice? Thanks a lot!

githubharald commented 4 years ago

did you call beam search without language model (which I would suggest, i.e. lm=None)?

If you have a neural network output for which best path performs better than beam search, you can provide it here and I can have a short look into it. Provide the relevant information: mat, classes, and beamWidth. And the expected output, of course. You can dump the matrix mat with np.save() function.

githubharald commented 4 years ago

closing because of inactivity