Result of test.py file about looking-to-listen-at-the-cocktail-party

[JusperLee]

You need to see for yourself if there is a problem with the preprocessed files generated.

[83344rushikesh]

@JusperLee your GPU size was in TB or GB ?

[JusperLee]

one gpu was in 64GB

[83344rushikesh]

@JusperLee okay..All the preprocessing has been done according to your readme only. Can you give possible solutions to this ?

[JusperLee]

Due to the different experimental environments, I think you need to carefully check your code, including testing and preprocessing. This code did not show the above problems during my experiment.

[83344rushikesh]

@JusperLee what was the accuracy of your audio visual model ?

[JusperLee]

Since I can't get the complete data set, the results I get are not as good as the values in the paper. 83344rushikesh <[email protected]> 于2020年3月30日周一 上午12:00写道：

…

@JusperLee <https://github.com/JusperLee> what was the accuracy of your audio visual model ? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#6 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AIB5NQQE7O5T2V3EGFBKD2TRJ5WDJANCNFSM4LV3BHHQ> .

[83344rushikesh]

@JusperLee Do you have your personal paper available for this in english ?

[JusperLee]

Looking to Listen at the Cocktail Party: A Speaker-Independent Audio-Visual Model for Speech Separation

[83344rushikesh]

@JusperLee How to use this model on a new video? Can you share that code?

[JusperLee]

You just need to change the input video.

[83344rushikesh]

@JusperLee For that new video we again need to do the face embedding like all the preprocessing from start right ?

[JusperLee]

Yes

[83344rushikesh]

@JusperLee For how many speakers it will work?

[83344rushikesh]

@JusperLee I'm getting following error while running the test.py file
Traceback (most recent call last):
File "test.py", line 92, in
T = utils.fast_istft(F,power=False)
File "/home/lenovo/Downloads/Looking-to-Listen-at-the-Cocktail-Party-master/utils.py", line 75, in fast_istft
data = istft(real_imag_shrink(data))
File "/home/lenovo/Downloads/Looking-to-Listen-at-the-Cocktail-Party-master/utils.py", line 30, in istft
Total[start:end] = Total[start:end] + data[i, :] * windows
ValueError: operands could not be broadcast together with shapes (257,) (512,)

[JusperLee]

1. This model does not limit the number of people mixed.
2. The different dimensions may be a problem with your own operation, which I cannot help.

[83344rushikesh]

@JusperLee At the place of 257 there should be 512 right?

[JusperLee]

I think you can check this out. These questions are very basic.