diff --git a/CHANGELOG.md b/CHANGELOG.md index eaba738a5..0169d74c8 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,6 +1,6 @@ # Changelog -## Version 2.1 (2022-11-xx) +## Version 2.1.1 (2022-10-27) - BREAKING(pipeline): rewrite speaker diarization pipeline - feat(pipeline): add option to optimize for DER variant diff --git a/README.md b/README.md index 15f2a9fe6..388dfb4b5 100644 --- a/README.md +++ b/README.md @@ -103,17 +103,17 @@ pip install -qq https://github.com/pyannote/pyannote-audio/archive/develop.zip Out of the box, `pyannote.audio` default speaker diarization [pipeline](https://hf.co/pyannote/speaker-diarization) is expected to be much better (and faster) in v2.x than in v1.1. Those numbers are diarization error rates (in %) -| Dataset \ Version | v1.1 | v2.0 | v2.1 (finetuned) | -| ---------------------- | ---- | ---- | ---------------- | -| AISHELL-4 | - | 14.6 | 14.1 (14.5) | -| AliMeeting (channel 1) | - | - | 27.4 (23.8) | -| AMI (IHM) | 29.7 | 18.2 | 18.9 (18.5) | -| AMI (SDM) | - | 29.0 | 27.1 (22.2) | -| CALLHOME (part2) | - | 30.2 | 32.4 (29.3) | -| DIHARD 3 (full) | 29.2 | 21.0 | 26.9 (21.9) | -| VoxConverse (v0.3) | 21.5 | 12.6 | 11.2 (10.7) | -| REPERE (phase2) | - | 12.6 | 8.2 ( 8.3) | -| This American Life | - | - | 20.8 (15.2) | +| Dataset \ Version | v1.1 | v2.0 | v2.1.1 (finetuned) | +| ---------------------- | ---- | ---- | ------------------ | +| AISHELL-4 | - | 14.6 | 14.1 (14.5) | +| AliMeeting (channel 1) | - | - | 27.4 (23.8) | +| AMI (IHM) | 29.7 | 18.2 | 18.9 (18.5) | +| AMI (SDM) | - | 29.0 | 27.1 (22.2) | +| CALLHOME (part2) | - | 30.2 | 32.4 (29.3) | +| DIHARD 3 (full) | 29.2 | 21.0 | 26.9 (21.9) | +| VoxConverse (v0.3) | 21.5 | 12.6 | 11.2 (10.7) | +| REPERE (phase2) | - | 12.6 | 8.2 ( 8.3) | +| This American Life | - | - | 20.8 (15.2) | ## Citations diff --git a/version.txt b/version.txt index 879b416e6..3e3c2f1e5 100644 --- a/version.txt +++ b/version.txt @@ -1 +1 @@ -2.1 +2.1.1