将some_model.zip中的some_model文件夹拖放至Dataset-tool根目录即可（结构层次为Dataset-tool/some_model/model.onnx）,fbl_model文件夹包括model.onnx文件。

也可使用其他版本的some onnx:
input:
waveform: float32[1,n_samples]
output:
note_midi: float32[1,n_notes]
note_rest: boolean[1,n_notes]
note_dur: float32[1,n_notes]

fix bug

minlabel的json新增key isCheck，用于识别是否人工检查过

fix FBL bug

Currently, FoxBreatheLabeler only supports annotating breathing using TextGrid files output from SOFA(i.e. overlaying new "AP" annotations on intervals already marked as "SP").

Warning: When using Sofa to output tg files, the parameter “--ap_detector NoneAPDetector” must be added.

For details, please refer to the How to use project 0 in the Python version.

The Python version can use CUDA acceleration and requires self installation of the Torch environment.

Download Fbl Model

Python version

FoxBreatheLabeler目前仅支持基于SOFA输出的TextGrid文件标注呼吸(即在已标为“SP”的区间覆盖上新的“AP”标注)。

将fbl_model.zip中的fbl_model文件夹拖放至Dataset-tool根目录即可（结构层次为Dataset-tool/fbl_model/model.onnx）,fbl_model文件夹包括model.onnx、config.yaml两个文件。

ASR results can be saved as pinyin.