Question about make_h5.py about mist_vad HOT 4 CLOSED

Hi @Jupiter-hzh, as described in the h5py documentation, the data parameter in create_dataset should be a NumPy array. Therefore, I fixed this error by converting tmp_frames, which is a list, to a NumPy array before passing it to create_dataset. In any case, I would like to hear from the authors if this is the correct way to handle the problem.

from mist_vad.

The target of using h5py is just reducing the indexing time in training stage. If you want to test it only, you can rewrite the datasets.py to read videos directly.

from mist_vad.

Hi @Jupiter-hzh, as described in the h5py documentation, the data parameter in create_dataset should be a NumPy array. Therefore, I fixed this error by converting tmp_frames, which is a list, to a NumPy array before passing it to create_dataset. In any case, I would like to hear from the authors if this is the correct way to handle the problem.

Thanks. For anyone has the same error, here is the snipet:

            tmp_frames = np.asarray(tmp_frames)
            h.create_dataset(key,data=tmp_frames,chunks=True)

from mist_vad.

I have updated the codes of this part, thanks @chuong98 @luca-zanella-dvl

from mist_vad.