Value error in nnabla on running x-umx on Rpi4, Raspberry Pi OS about ai-research-code

[RajivX]

Yes, i could solve it successfully. I installed the latest version of nnabla, ver 1.17 as of today,
root@raspberrypi:/home/pi/x-umx/results# pip3 freeze | grep 'nnabla'
nnabla==1.17.0.dev1

I ported x-umx successfully on my Rpi4, 8 GB. Yay...!!
This is the first succesful porting & running of x-umx on Rpi4.
I have run x-umx on a speech capture recorded with a MIC.
The speech capture is 16kHz, Stereo, S16_LE format wav.

After running this,
root@raspberrypi:/home/pi/x-umx# python3 test.py --inputs ../Music/test_16k_S16_LE_stereo.wav --context cpu --model /home/pi/x-umx/x-umx.h5 --outdir /home/pi/x-umx/results

2021-01-31 08:16:34,465 [nnabla][INFO]: Initializing CPU extension...

Four outputs are generated inside 'results' folder(you can keep any other folder path as you pass in --outdir):

root@raspberrypi:/home/pi/x-umx/results# ls
bass.wav drums.wav other.wav vocals.wav

I played each of them. It there was a 1 kHz background noise, which was there in the recording capture.
It has been eliminated at some places of the track & minimized throughout the track.
SO, yes, x-umx can be used as Speech Noise removal.

Spectrogram analysis to be done further for better analysis further. I will be attaching my output & input
for everyone here.

Thanks a lot, x-umx team from SONY.

Regards,
Rajiv.

[RajivX]

Attaching the INPUT recording capture done on Rpi4 using AMIC Soundcard(4-MIC KIT, AC108 ADC),[test_16k_S16_LE_stereo.wav],

And the OUTPUTS generated on Rpi4, 8GB after running x-umx,[bass.wav, drums.wav, other.wav, vocals.wav]
Please download the attached zip, [results_and_input.zip].

results_and_input.zip

Please analyze.
Regards,
Rajiv.

[RajivX]

Please close the issue, as i have successfully resolved it.

Rajiv.

[TomonobuTsujikawa]

Thank you for running x-umx on Rpi4 environment!
We would like to support Rasberrry Pi environment in the near future!