tensorflow_end2end_speech_recognition icon indicating copy to clipboard operation
tensorflow_end2end_speech_recognition copied to clipboard

End-to-End speech recognition implementation base on TensorFlow (CTC, Attention, and MTL training)

Results 11 tensorflow_end2end_speech_recognition issues
Sort by recently updated
recently updated
newest added

Hello, Thanks for your work! How do I run the pre-trained models for speech recognition? Thanks,

If I want to train this project for Bangla language then what files are needed to change? Please give me the instructions.

Hi, This is very interesting project. I have a question on the CTC beam search (with charlm) you're currently implementing. Are you planning to use RNN based character language model...

It seems that attention layer type 'location' should actually be 'hybrid' [1,2]. [1] Kim, Suyoun, Takaaki Hori, and Shinji Watanabe. "Joint CTC-attention based end-to-end speech recognition using multi-task learning." Acoustics,...

This project's input label path and asr_preprocessing's feature path do not match? Is this project still under maintenance?

In the class AttentionLayer (models/attention/decoders/attention_layer.py), the initial parameter does not include "sigmoid_smoothing". But, in models/attention/attention_seq2seq.py, it calls AttentionLayer with sigmoid_smoothing=self.sigmoid_smoothing (line 338). Which file should I modify? Thx!

I'm refactoring attention-based models, so they do not work now! Please wait for a few days.

Hi, just wanna check out when will this repo update? It would be better if some baseline speech recognition results could be provided. Last time I tried, I found there...

I got an error as ---------------- if l == padded_value: ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all() --------------- I...

I found many places using padded_value=train_data.padded_value However, the Dataset object does not include padded_value Only in load_dataset_ctc.py is different: self.padded_value = -1 So, what is the value of padded_value should...