Sry but How to get the correctly formatted hdf5 file from the downloaded dataset?

[betairylia]

Sorry for a silly question.

I found that your code reads a hdf5 file as a input, but there're only many frames and audios in your public dataset. Since hdf5 file contains lots of formatted data structures, I think there must be some scripts to convert the dataset folder to the target hdf5 file for further training. However, I could not find such file, and also cannot figure out how to get that correct hdf5 file. May I ask that what should I do in order to get the correct file?

Thanks !

[betairylia]

Well, where can I find the dataset? or is it not supposed to provide the dataset? Thanks.

[rhgao]

The dataset used is AudioSet (https://research.google.com/audioset/ontology/index.html), and you could choose the classes that you want to use and download the corresponding video clips. The categories used in the paper are listed in Supp.

[rhgao]

To generate hdf5 file, you could use scripts similar to below:
import h5py
h5f = h5py.File('train.h5', 'w')
h5f.create_dataset('bases', data=bases[0:numOfTrain])
h5f.create_dataset('labels', data=labels[0:numOfTrain])
h5f.close()

"bases" is a numpy array that stores the bases extracted from NMF, and "labels" stores the labels predicted from visual classifier or ground-truth labels. A good NMF implementation is: http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.NMF.html

[betairylia]

To generate hdf5 file, you could use scripts similar to below:
import h5py
h5f = h5py.File('train.h5', 'w')
h5f.create_dataset('bases', data=bases[0:numOfTrain])
h5f.create_dataset('labels', data=labels[0:numOfTrain])
h5f.close()

"bases" is a numpy array that stores the bases extracted from NMF, and "labels" stores the labels predicted from visual classifier or ground-truth labels. A good NMF implementation is: http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.NMF.html

Thanks for the information! You have done great work!

[betairylia]

To generate hdf5 file, you could use scripts similar to below:
import h5py
h5f = h5py.File('train.h5', 'w')
h5f.create_dataset('bases', data=bases[0:numOfTrain])
h5f.create_dataset('labels', data=labels[0:numOfTrain])
h5f.close()

"bases" is a numpy array that stores the bases extracted from NMF, and "labels" stores the labels predicted from visual classifier or ground-truth labels. A good NMF implementation is: http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.NMF.html

Sorry for disturbing again, should it be:
bases array shaped [numOfTrain, M (#basis vectors), vector_dim (of basis vector)], and
labels array shaped [numOfTrain, label_dim] ?

[rhgao]

In my case, bases is just a list of the paths that stores the actual bases, and labels are the corresponding label indexes in the same order. Then I load the bases in the data loader through the corresponding path.

[betairylia]

Thank you very much for those information! (Sorry for the late reply.)