About extracted feature

[WoshiBoluo]

Hello， your work is excellent! Could you please make public the extracted feature on Google Drive？

[yangdongchao]

Hello， your work is excellent! Could you please make public the extracted feature on Google Drive？

I realy want to do this, but there is no space in my google drive. So I cannot share it on Google Drive now. Actually, you can use the baseline of DCASE2021 task5 to extract the feature. I will try to share it on BaiDu cloud desk in the next few days.

[WoshiBoluo]

Hello， your work is excellent! Could you please make public the extracted feature on Google Drive？

I realy want to do this, but there is no space in my google drive. So I cannot share it on Google Drive now. Actually, you can use the baseline of DCASE2021 task5 to extract the feature. I will try to share it on BaiDu cloud desk in the next few days.

Thank you very much. After I used the official feature_extract, I found that the Mel_train.h5 of the train is only 152MB. Is this correct?

[yangdongchao]

Hello， your work is excellent! Could you please make public the extracted feature on Google Drive？

I realy want to do this, but there is no space in my google drive. So I cannot share it on Google Drive now. Actually, you can use the baseline of DCASE2021 task5 to extract the feature. I will try to share it on BaiDu cloud desk in the next few days.

Thank you very much. After I used the official feature_extract, I found that the Mel_train.h5 of the train is only 152MB. Is this correct?

It is right. The training data is very small, but the validation data is much larger.

[WoshiBoluo]

Thanks for your reply, I have another question about the evaluation set, I found that the official evaluation set provided by the .csv file only contains 5 POS labels per file, where do I get all the labels to evaluate my model?

[yangdongchao]

Thanks for your reply, I have another question about the evaluation set, I found that the official evaluation set provided by the .csv file only contains 5 POS labels per file, where do I get all the labels to evaluate my model?

The evalution set label is only can be used during contest.

[liushenme]

Hi, your project is very interesting. I see in your paper you say the feature you used is STFT followed by a Mel-scaled filter bank on perceptually weighted spectrograms. Is it same as PCEN baseline of the task used ?

[yangdongchao]

Yes it is the same as the baseline.

[liushenme]

Yes it is the same as the baseline.

I see. Thank you.

[Noumanijaz744]

Hello， your work is excellent! Could you please make public the extracted feature on Google Drive？

I realy want to do this, but there is no space in my google drive. So I cannot share it on Google Drive now. Actually, you can use the baseline of DCASE2021 task5 to extract the feature. I will try to share it on BaiDu cloud desk in the next few days.

Thank you very much. After I used the official feature_extract, I found that the Mel_train.h5 of the train is only 152MB. Is this correct?

Hello Sir @yangdongchao the official development datasets contains training set of 14 hours and 20 mins of audio recording while the validation set contains only 5 hours of audio recording so how the Mel_train.h5 is too smaller than the Mel_Eval.h5 .please if you can clear my question .thank you