Giters

ymhzyj / UMMAFormer

[ACM MM'23] UMMAFormer: A Universal Multimodal-adaptive Transformer Framework For Temporal Forgery Localization

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

AudioNTT2020Feature

Myblackcat0216 opened this issue · comments

commented

Hello, what is the AudioNTT2020Feature from byol_a.models import AudioNTT2020Feature in byola_extract_lavdf.py
Traceback (most recent call last):
File "/home/UMMAFormer-main/bylo-a/byola_extract_lavdf.py", line 3, in
from byol_a.models import AudioNTT2020Feature
ImportError: cannot import name 'AudioNTT2020Feature' from 'byol_a.models' (/home/UMMAFormer-main/bylo-a/byol_a/models.py)
Another question is what parameters need to be changed when extracting audio features using byol-a？
Thank you for your excellent work and look forward to your reply

commented

Thank you for your attention. If you want to use BYOL-A to extract features, you need to use the code available at https://github.com/nttcslab/byol-a. What I provided is just a basic example of calling the relevant code.

commented

@ymhzyj Dear author, thank you for using BYOL-A and excuse me for cutting in.
Is AudioNTT2020Feature the same as AudioNTT2020Task6? That could be the answer to the question.
https://github.com/nttcslab/byol-a/blob/master/byol_a/models.py#L48

@Myblackcat0216 I think you can use AudioNTT2020Task6 if you want to extract audio features for each time frame.
If you want a single feature vector for an audio clip, please try AudioNTT2020 instead.

commented

Thank you very much for your help. It solved my problem