rhgao / deep-miml-network Goto Github PK

Dear Mr. Gao
Thank you so much for the great work. However, I met some problems when I implemented this code.
As described in you article, "For the visual frames, we use an ImageNet pre-trained ResNet-152 network [34] to make object category predictions, and we max-pool over predictions of all frames to obtain a video-level prediction. The top labels (with class probability larger than a threshold = 0.3) are used as weak \labels" for the unlabeled video."
However, when I use the pre-trained-152 network, I can get the only one category prediction lager than the threshold. How can I get multi-labels through the pre-trained-152 network.
Should I train a object detection network or a multi-classes multi-labels network or some other solutions. Thank you for your assistance
Best regards!

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 !

hi, I want to use this model to test some data, can you provide the well trained weight file for us? thanks

Hi rhgao,

Thanks for releasing your code. When I used your network to seperate audio sources, to get some .wav files as output, I realized a NMF algorithm estimating H with the W fixed by myself. But I found my program didn‘t seem to converge. So I want to know how did you compute H with W fixed? Do you know of any related libraries? I didn't found it in sklearn you mentioned in #1. Or did you realize it by yourself? And could you share your code?

Thank you very much! Waiting for your reply.