jhammarstedt / SLN-on-Sound

Re-implementation of the Noise Against Noise: Stochastic Label Noise helps combat inherent label noise

SLN with noise

A reimplementation of the Semi-supervised Learning paper on Stochastic Label Noise on the FSD50K sound dataset.

Credits

Sound data preprocessing and networks are credited to: https://github.com/SarthakYadav/fsd50k-pytorch

Some edits and workarounds were added to the code to make it work as intended in this scenario

Wide ResNet structure similar to the original implementation: https://github.com/chenpf1025/SLN

Authors:

• Johan Hammarstedt, Jhammarstedt
• Matej Sestak, Sestys
• Adam Wiker , AdamWiker