Speech Extensions to Fairseq
summary
Fairseq now includes extensions that add end-to-end speech and audio recognition pipelines: dataset readers and feature extraction, encoder/decoder modules wired for raw audio, training loops with sequence losses and beam-search decoding, and integration points for export to production runtimes. The demo shows the repo accepting audio files, training models directly from waveforms, and producing transcriptions end-to-end, enabling faster prototyping and a clear path to deploy models built and validated inside the Fairseq codebase.
AI-written summary · about
related talks
31m“Sorry, Could you repeat that again?” - Speech Recognition with Python
40mGathering Insights from Audio Data
25mLooking at Sound
16mThe End of Awkward AI Transcriptions
31mOff-the-shelf HuggingFace models for audio deepfake detection
28mHow to use Deep Learning to process audio data and classify sounds.
20mVoice fingerprinting & recognition
06mSensing to Hear: Speech Enhancement for Mobile Devices Using Acoustic Signals