Code

C++ API for speaker diarization#1396

Code

Merged

csukuangfj merged 17 commits intok2-fsa:masterk2-fsa/sherpa-onnx:masterfrom csukuangfj:speaker-diarization-cppcsukuangfj/sherpa-onnx:speaker-diarization-cppCopy head branch name to clipboardOct 9, 2024

+1,652-108

Commits

Commits on Oct 7, 2024

• Begin to add C++ implementation for offline speaker diarization.

  csukuangfj

  committed
  a30ae88

  Copy full SHA for a30ae88

• add config for speaker diarization

  csukuangfj

  committed
  dd65064

  Copy full SHA for dd65064

• begin to add pyannote segmentation model

  csukuangfj

  committed
  23fd7ec

  Copy full SHA for 23fd7ec

Commits on Oct 8, 2024

• Run the speaker segmentation model

  csukuangfj

  committed
  d0e0f99

  Copy full SHA for d0e0f99

• refactoring

  csukuangfj

  committed
  b9bcbfa

  Copy full SHA for b9bcbfa

• Convert segmentations to multi-labels.

  csukuangfj

  committed
  1cf26d4

  Copy full SHA for 1cf26d4

• Compute speaker count per frame

  csukuangfj

  committed
  42e1eff

  Copy full SHA for 42e1eff

• Get chunk-speaker pair and its sample indexes

  csukuangfj

  committed
  9b4f691

  Copy full SHA for 9b4f691

• compute embeddings

  csukuangfj

  committed
  b91a4be

  Copy full SHA for b91a4be

• Add clustering

  csukuangfj

  committed
  9931a26

  Copy full SHA for 9931a26

• first working version

  csukuangfj

  committed
  06c0d45

  Copy full SHA for 06c0d45

• refactor

  csukuangfj

  committed
  ee263e6

  Copy full SHA for ee263e6

• change optset to 13

  csukuangfj

  committed
  e972b2e

  Copy full SHA for e972b2e

• use onnxruntime 1.16.3 and onnx 1.15.0 to export pyannote models

  csukuangfj

  committed
  24ec150

  Copy full SHA for 24ec150

• first working version

  csukuangfj

  committed
  613df17

  Copy full SHA for 613df17

Commits on Oct 9, 2024

• Add CI tests

  csukuangfj

  committed
  b6dea7e

  Copy full SHA for b6dea7e

• fix typos

  csukuangfj

  committed
  1eb1a50

  Copy full SHA for 1eb1a50