TeaPoly/CTC-OptimizedLoss
Computes the MWER (minimum WER) Loss with CTC beam search. Knowledge distillation for CTC loss.
This project provides specialized loss functions for Connectionist Temporal Classification (CTC) models, which are commonly used in speech recognition. It takes raw CTC outputs (logits) and target labels, then calculates an optimized loss that can improve the accuracy of speech-to-text systems. Speech recognition engineers and researchers would use this to fine-tune their CTC-based models.
No commits in the last 6 months.
Use this if you are developing or training a speech recognition model using CTC and want to apply advanced loss optimization techniques like Minimum Word Error Rate (MWER) or knowledge distillation to improve performance.
Not ideal if you are a general practitioner looking for an out-of-the-box speech recognition solution or if you are not working directly with CTC model training.
Stars
59
Forks
11
Language
Python
License
—
Category
Last pushed
Sep 06, 2023
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/voice-ai/TeaPoly/CTC-OptimizedLoss"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
githubharald/CTCDecoder
Connectionist Temporal Classification (CTC) decoding algorithms: best path, beam search, lexicon...
githubharald/CTCWordBeamSearch
Connectionist Temporal Classification (CTC) decoder with dictionary and language model.
nl8590687/ASRT_SpeechRecognition
A Deep-Learning-Based Chinese Speech Recognition System 基于深度学习的中文语音识别系统
athena-team/athena
an open-source implementation of sequence-to-sequence based speech processing engine
hirofumi0810/tensorflow_end2end_speech_recognition
End-to-End speech recognition implementation base on TensorFlow (CTC, Attention, and MTL training)