search

tensorflow / models

Models and examples built with TensorFlow
Other
76.97k stars 45.79k forks source link

audioset hyper-parameter problem #3479

Closed guoxiaolu closed 6 years ago

guoxiaolu commented 6 years ago

I have noticed that given the wav file, audioset will output sec9664 examples_batch and final output sec*128 embedding output. I want to get a larger and fixed embedding output such as 400, but I don't know how to change the hyper-parameter in vggish-params.py. I'm new to wave problems and can you give me some suggestion? @plakal and @dpwe Thank you very much.

plakal commented 6 years ago

Please use the mailing list https://groups.google.com/forum/#!forum/audioset-users for these kinds of questions and discussion. Use the issue tracker for specific feature request or bug reports or code contributions.

Closing.

ucas010 commented 4 years ago

大佬，我也是遇到这个问题了，咋办啊？ 如果说不同音频长度，那么其shape肯定不同，(batch,128),这怎么能作为最后的embedding呢？有点懵逼啊，大佬现在有啥解决的方法没啊？请帮帮我吧，谢谢