LASAFT-Net-v2

Woosung Choi, Yeong-Seok Jeong, Jinsung Kim, Jaehwa Chung, Soonyoung Jung, and Joshua D. Reiss

Welcome to the tutorial webpage of LASAFT-Net-v2, which is a conditioned source separation model.

It outperforms our previous version, called LaSAFT-Net-v1 [CKCJ21], equipping LASAFT block-v2.

While the existing method [CKCJ21] only cares about the symbolic relationships between the target source symbol and latent sources, ignoring audio content, our new approach also considers audio content with listening mechanisms.

Below is the experimental results of LASAFT-Net-V2.

Experimental Results

Musdb 18 (No extra dataset)

model

conditioned?

vocals

drums

bass

other

AVG

Demucs

X

6.84

6.86

7.01

4.42

6.28

D3Net

X

7.24

7.01

5.25

4.53

6.01

Meta-TasNet

O

6.40

5.91

5.58

4.19

5.52

AMSS-Net

O

6.78

5.92

5.10

4.51

5.58

LaSAFT-Net-v1

O

7.33

5.68

5.63

4.87

5.88

LASAFT-Net-v2

O

7.57

6.13

5.28

4.87

5.96

LASAFT-Net-v2 (updated)

O

7.43±0.09

6.23±0.05

5.28±0.19

4.89±0.05

5.99±0.03

MDX Challenge (Leaderboard A)

model

conditioned?

vocals

drums

bass

other

Song

Demucs++

X

7.968

8.037

8.115

5.193

7.328

KUILAB-MDX-Net

X

8.901

7.173

7.232

5.636

7.236

Kazane Team

X

7.686

7.018

6.993

4.901

6.649

LASAFT-Net-v2.0

O

7.354

5.996

5.894

4.595

5.960

LaSAFT-Net-v1.2

O

7.275

5.935

5.823

4.557

5.897

Demucs48-HQ

X

6.496

6.509

6.470

4.018

5.873

LaSAFT-Net-v1.1

O

6.685

5.272

5.498

4.121

5.394

XUMXPredictor

X

6.341

5.807

5.615

3.722

5.372

UMXPredictor

X

5.999

5.504

5.357

3.309

5.042

LaSAFT-Net-v1.1 is also known as lightsaft-net

In this tutorial, we explain how to use, train and evaluate our model.