Quality Reduction in v3 about vocal-remover

[404000]

Anjok07 in version 3.0.1 tsurumuso wrote than if you want to improve the vocal removal quality with TTA, use the --tta option
Does this tta improve accompaniment quality?

[Anjok07]

Anjok07 in version 3.0.1 tsurumuso wrote than if you want to improve the vocal removal quality with TTA, use the --tta option
Does this tta improve accompaniment quality?

Not really. The quality of the accompaniment isn't even close to v2, even with the tta option on.

[404000]

How did you know that the quality of accompaniment decreased so much?

[tsurumeso]

Was this to make the vocal removal more aggressive? If so, what parameters need to be tweaked to bring the accompaniment quality back up to v2?

Probably so.
You can adjust the vocal removal level with the --max_reduction_rate option. By setting this to 0, the behavior is almost the same as v2.

[404000]

Tsurumeso Will it be like this?python inference.py --input path/to/an/audio/file --max_reduction_rate 0

[tsurumeso]

It's a training option. It will be like this:

python train.py --dataset path/to/dataset --max_reduction_rate 0 --mixup_rate 0.5 --gpu 0

[Anjok07]

Thank you for the feedback! I will try this

[tsurumeso]

Sorry, there was a bug in v3.0.3. Please use v3.0.4 as follows:

python train.py --dataset path/to/dataset --reduction_rate 0 --mixup_rate 0.5 --gpu 0