DCASE-REPO / dcase2018_baseline

DCASE 2018 Baseline systems
MIT License
128 stars 78 forks source link

Videos for these audios? #27

Closed HarshVardhanSh closed 5 years ago

HarshVardhanSh commented 5 years ago

I want to access the videos associated with these audios. I used the VideoIDs from these audio filenames but youtube says " the video is unavailable". Please help. Thanks

HarshVardhanSh commented 5 years ago

Nobody has not yet commented on this thread.

turpaultn commented 5 years ago

Some of the videos are now unavailable on youtube so it is impossible to get the video associated.

HarshVardhanSh commented 5 years ago

I guess all of the videos for weakly labelled audios are unavailable. BTW thanks for the confirmation.

turpaultn commented 5 years ago

No most of them are available: Examples of the two first ones: https://www.youtube.com/watch?v=KK227gPpRn4 https://www.youtube.com/watch?v=-6p32k2WUCs

HarshVardhanSh commented 5 years ago

I couldn't find the videos you mentioned in the dataset. I guess we are talking about different datasets. I am referring to task 4 of the competition. Following are the sample audios from the weak folder of the audio dataset in task 4: Y-_A4yJmxFdM_30.000_40.000 Y_CZzm6jUxbo_30.000_40.000 Y_KNIJQJ-CHA_30.000_40.000

As written the dataset description, Y-_A4yJmxFdM will be the VideoID for the video associated with the first audio. So I tried using these videoIDs for almost every audio in this folder and for all of them, the youtube says "Video unavailable".

Let me know where am I failing ?

turpaultn commented 5 years ago

The Y at the beginning is audioset convention but shouldn't be put in the ID on youtube. So youtube IDs: Y-_A4yJmxFdM_30.000_40.000 --> -_A4yJmxFdM Y_CZzm6jUxbo_30.000_40.000 --> _CZzm6jUxbo Y_KNIJQJ-CHA_30.000_40.000 --> _KNIJQJ-CHA

You would find it in dataset/download_data.py (line 41): query_id = filename[1:12]

Then line 60: https://www.youtube.com/watch?v={query_id}

HarshVardhanSh commented 5 years ago

Gotcha! Thanks so much!