search

Jungjee / DcaseNet

Author's repository for reproducing DcaseNet, an integrated pre-trained DNN that performs acoustic scene classification, audio tagging, and sound event detection. Implemented using PyTorch.
MIT License
40 stars 7 forks source link

If you can please add links to the datasets #7

Closed Danzip closed 3 years ago

Danzip commented 3 years ago

Can you also add in the readme how you need the directory tree to look like? Those datasets need to be extracted and they have a certain directory tree can you add in the readme an example of how it should like?

Jungjee commented 3 years ago

Hi, filetree is available here

Jungjee commented 3 years ago

all three datasets should be under one root folder :)

Danzip commented 3 years ago

thank you very much

Danzip commented 3 years ago

I'm sorry but i looked in the SED file-tree and i cant figure out what all those files are the .npy and the. csv all i have from the site is the wav files can you add a link? or explain? i am getting the error in the pic Screenshot from 2021-02-16 16-57-33 i put the wav files in the feat_label folder like so /hdd/data/dcase_2020_task3_extracted/feat_label/foa_dev foa_dev contains all of the wavs

Jungjee commented 3 years ago

Hi, can you check below things? you can ignore .npy files since they belong to different experiments. I should remove them in this repo. The basic is that 'h5' files are made using the original 'wav' files.

  1. if files with 'h5' extension are created under 'log_mel_spec_label'?
  2. whether this line is successfully executed

In my opinion, error occurred because the preprocessing wasn't executed correctly