search

xmos / fwk_voice

Voice Framework
Other
11 stars 19 forks source link

Port py_ic v2.3.0 to fwk_voice #369

Closed shuchitak closed 2 years ago

shuchitak commented 2 years ago
shuchitak commented 2 years ago

Avona full pipeline keyword test results compared with lib_audio_pipelines

Alt-arch Alt_arch_DUT_Audio___XMOS_DUT1_Env_Audio___Clean Alt_arch_DUT_Audio___XMOS_DUT1_Env_Audio___Noise1 Alt_arch_Env_Audio___Noise1 Alt_arch_Env_Audio___Noise2

shuchitak commented 2 years ago

Avona full pipeline keyword test results compared with lib_audio_pipelines

Prev-arch

Prev_arch_DUT_Audio___XMOS_DUT1_Env_Audio___Clean Prev_arch_DUT_Audio___XMOS_DUT1_Env_Audio___Noise1 Prev_arch_Env_Audio___Noise1 Prev_arch_Env_Audio___Noise2

shuchitak commented 2 years ago

Avona pipeline results compared with the python pipeline in prev_arch mode with Initial delay estimation using the ADEC disabled.

Python_DUT_Audio___XMOS_DUT1_Env_Audio___Clean Python_DUT_Audio___XMOS_DUT1_Env_Audio___Noise1 Python_Env_Audio___Noise1 Python_Env_Audio___Noise2

mbanth commented 2 years ago

@uvvpavel or @shuchitak, what additional work needs to occur to resolve this issue?

shuchitak commented 2 years ago

Hi @mbanth, the PR for this is merged and I'll close the issue now. We've published the kwd test results and if any work is ever done to address a drop in performance for prev_arch DUT streams, that will be addressed in a separate issue.