Kuvaus
This is a .zip file containing the SDR, SIR, and SAR results (per file) of the MaD TwinNet, evaluated on the Test subset of the Demixing Secret Dataset (DSD). Each file in the .zip archive, is a file containing the evaluation results of the MaD TwinNet method. The results have been serialized (i.e. saved to the hard disk) using the Python 3.6v pickle package and protocol=2 The order of the results for each metric, is the same as the order of the evaluation files at the DSD. This means that each result file, contains a list of arrays (the arrays are bumpy ndarrays). Note bold: The LICENSE of the code of the MaD TwinNet applies also to these results.
| Koska saatavilla | 2 helmik. 2018 |
|---|---|
| Julkaisija | Zenodo |
Field of science, Statistics Finland
- 113 Tietojenkäsittely ja informaatiotieteet
Tutkimustuotos
- 1 Konferenssiartikkeli
-
MaD TwinNet: Masker-Denoiser Architecture with Twin Networks for Monaural Sound Source Separation
Drossos, K., Mimilakis, S. I., Serdyuk, D., Schuller, G., Virtanen, T. & Bengio, Y., 10 heinäk. 2018, 2018 International Joint Conference on Neural Networks (IJCNN). IEEETutkimustuotos: Konferenssiartikkeli › Tieteellinen › vertaisarvioitu
19 Sitaatiot (Scopus)
Tietoaineistot
-
MaD TwinNet Pre-trained Weights
Drossos, K. (Creator), Mimilakis, S. I. (Creator), Serdyuk, D. (Creator), Schuller, G. (Creator), Virtanen, T. (Creator) & Bengio, Y. (Creator), Zenodo, 2 helmik. 2018
DOI - pysyväislinkki: 10.5281/zenodo.1164592
Tietoaineisto: Dataset
Siteeraa tätä
- DataSetCite