2 ms·
Note that Mozilla's "Project DeepSpeech" is just a training and recognition engine, and really an implementation of a specific design for LVCSR. A full recognit
by gok 8y ago
Note that Mozilla's "Project DeepSpeech" is just a training and recognition engine, and really an implementation of a specific design for LVCSR. A full recognition system requires statistical models, generally trained from very large sources of data, which you have to bring yourself.
- singularity2001 8y agodon't they provide (pre)trained models? someone on github does.
- stephensonsco 8y agoPretrained models are very nice to spin up a system and start using it. You need that because training the model is so hard. But, the pretrained models are by no means a good general model. They are trained with narrow params (like #/size of datasets), so the ability to generalize is very low. It's not uncommon to train a system like that to be 5% WER on Switchboard (think MS/IBM), then have that same system perform at 40% WER on other audio.
- ephes 8y agoThe data is also included (for english, data for other languages is being currently collected). Just search for "common voice".