Upload folder using huggingface_hub

README.md

@@ -326,10 +326,8 @@ speech, counted as speech in another language.
 | Multilingual LibriSpeech | [facebook/multilingual_librispeech](https://huggingface.co/datasets/facebook/multilingual_librispeech) `2e83e618` | test, de/fr/es/it/nl/pt | CC BY 4.0 | speech in other languages | 2.00 | 1407 |
 | MUSAN music | [corypaik/musan](https://huggingface.co/datasets/corypaik/musan) `76f9882c` | music | CC BY 4.0, per-track terms | music | 2.05 | 0 |
 | MUSAN noise | [corypaik/musan](https://huggingface.co/datasets/corypaik/musan) `76f9882c` | noise | CC BY 4.0, per-clip terms | household and outdoor sounds | 0.83 | 0 |
-| DEMAND | [philgzl/demand](https://huggingface.co/datasets/philgzl/demand) `29f46e61` | 16k | CC BY 4.0 | household and outdoor sounds | 0.40 | 0 |
-| Public-domain sounds of pdsounds.org | [TigreGotico/public_domain_sounds_3secs](https://huggingface.co/datasets/TigreGotico/public_domain_sounds_3secs) `f47136f9` | all | public domain | household and outdoor sounds | 0.40 | 0 |
-| ESC-50 | [TigreGotico/ESC-50](https://huggingface.co/datasets/TigreGotico/ESC-50) `f4368b7d` | all | CC BY-NC 3.0 (non-commercial) | household and outdoor sounds | 0.20 | 0 |
-| FSD50K, clips with no human voice | [Fhrozen/FSD50k](https://huggingface.co/datasets/Fhrozen/FSD50k) `ccf1acaf` | eval | CC0, CC BY or CC BY-NC per clip (non-commercial) | household and outdoor sounds | 0.20 | 0 |
+| DEMAND | [philgzl/demand](https://huggingface.co/datasets/philgzl/demand) `29f46e61` | 16k | CC BY 4.0 | household and outdoor sounds | 0.60 | 0 |
+| FSD50K, clips with no human voice | [Fhrozen/FSD50k](https://huggingface.co/datasets/Fhrozen/FSD50k) `ccf1acaf` | eval | CC0, CC BY or CC BY-NC per clip (non-commercial) | household and outdoor sounds | 0.60 | 0 |
 | Speech-MASSIVE pt-PT | [TigreGotico/speech_MASSIVE_pt-PT](https://huggingface.co/datasets/TigreGotico/speech_MASSIVE_pt-PT) `288e7b5f` | test | CC BY-NC-SA 4.0 (non-commercial) | read speech | 2.50 | 267 |
 | Basque Parliament, Basque segments | [gttsehu/basque_parliament_1](https://huggingface.co/datasets/gttsehu/basque_parliament_1) `46208b30` | test | CC0 | radio, parliament and talk recordings | 1.50 | 2 |
 | FLEURS de_de | [google/fleurs](https://huggingface.co/datasets/google/fleurs) `70bb2e84` | test | CC BY 4.0 | read speech | 2.50 | 2 |
@@ -359,7 +357,10 @@ These sources stay out:
   hard, so a rate on them does not say how often a detector wakes in a home.
 - **Sources a fighter trained on**: `TigreGotico/not-wake-words-speech-en`,
   AudioSet, LibriSpeech train, and the train splits of Common Voice and FLEURS.
-  Only the test splits of Common Voice and FLEURS are used.
+  Only the test splits of Common Voice and FLEURS are used. `TigreGotico/ESC-50`,
+  `TigreGotico/public_domain_sounds_3secs` and `TigreGotico/FMA_3secs` are in the
+  default negative and background-noise pools of the wakeforge trainer, so they stay
+  out too.
 - **Sources with no known licence**: `TigreGotico/ambient_noises` and
   `TigreGotico/building_106_kitchen_3secs`.
 - **`TigreGotico/NAR`**: it is for research use only, and it holds the word "stop".