update
This commit is contained in:
parent
fbe5bf1d21
commit
a29026cee4
23
README.md
23
README.md
@ -159,5 +159,28 @@ git clone https://www.modelscope.cn/iic/SenseVoiceSmall.git
|
|||||||
|
|
||||||
Undo
|
Undo
|
||||||
|
|
||||||
|
# Performance
|
||||||
|
|
||||||
|
## 语音识别效果
|
||||||
|
我们在开源基准数据集(包括 AISHELL-1、AISHELL-2、Wenetspeech、Librispeech和Common Voice)上比较了SenseVoice与Whisper的多语言语音识别性能和推理效率。在中文和粤语识别效果上,SenseVoice-Small模型具有明显的效果优势。
|
||||||
|
|
||||||
|
<p align="center">
|
||||||
|
<img src="fig/asr_results.png" alt="SenseVoice模型在开源测试集上的表现" width="500" />
|
||||||
|
</p>
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
## 情感识别效果
|
||||||
|
|
||||||
|
## 事件检测效果
|
||||||
|
|
||||||
|
|
||||||
|
## 推理效率
|
||||||
|
SenseVoice-small模型采用非自回归端到端架构,推理延迟极低。在参数量与Whisper-Small模型相当的情况下,比Whisper-Small模型推理速度快7倍,比Whisper-Large模型快17倍。同时SenseVoice-small模型在音频时长增加的情况下,推理耗时也无明显增加。
|
||||||
|
|
||||||
|
|
||||||
|
<p align="center">
|
||||||
|
<img src="fig/inference.png" alt="SenseVoice模型的推理效率" width="500" />
|
||||||
|
</p>
|
||||||
|
|
||||||
<p style="color: lightgrey;">如果您是本模型的贡献者,我们邀请您根据<a href="https://modelscope.cn/docs/ModelScope%E6%A8%A1%E5%9E%8B%E6%8E%A5%E5%85%A5%E6%B5%81%E7%A8%8B%E6%A6%82%E8%A7%88" style="color: lightgrey; text-decoration: underline;">模型贡献文档</a>,及时完善模型卡片内容。</p>
|
<p style="color: lightgrey;">如果您是本模型的贡献者,我们邀请您根据<a href="https://modelscope.cn/docs/ModelScope%E6%A8%A1%E5%9E%8B%E6%8E%A5%E5%85%A5%E6%B5%81%E7%A8%8B%E6%A6%82%E8%A7%88" style="color: lightgrey; text-decoration: underline;">模型贡献文档</a>,及时完善模型卡片内容。</p>
|
||||||
|
|||||||
BIN
fig/asr_results.png
Normal file
BIN
fig/asr_results.png
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 238 KiB |
BIN
fig/inference.png
Normal file
BIN
fig/inference.png
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 470 KiB |
Loading…
Reference in New Issue
Block a user