dcase20_task4
dcase20_task4 copied to clipboard
turpaultn
hello
I want to ask how much space do I need to download the dataset for your project? Do I need to run all the scripts to download the data just for run the MT_train?
I just want to run the baseline model but I don't know how to prepare my dataset.
I just want to run the baseline model but I don't know how to prepare my dataset.
It depends, if you just want to train the SED model without the Sound separation one, you can only download the SED part. You can check this script: 1_download_data.sh
You'll have some missing files with your "real_data" and will have to send a mail to [email protected] to get the other data
About the size of the dataset: 20G desed_soundbank (depends on the option) 23G desed_real
thanks for your reply. One more question Based on your your code in the 'data_generation' fold . In order to generate the dataset under the folder named "/dataset/audio/train/synthetic20/xxx" which is written in the main.py . Do I have to run the get_backgrouds_train in the 1_download_data.sh ? After download about 23G real data I want to know if I can skip this process. B
ok, I might figured it out . let me make my question clear. Can you tell me in which procedure or scripts to generate the folder '/dataset/audio/train/synthetic20/xxx' thanks!