Original Sound |
Reconstructed from the original mel-spectrogram using Griffin-Lim |
Reconstructed from the decoded mel-spectrogram using Griffin-Lim |
Reconstructed from the original mel-spectrogram using the sinusoidal signal reconstruction method |
Reconstructed from the decoded mel-spectrogram using the sinusoidal signal reconstruction method |
Baseline |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Dropout03Encoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Dropout03Decoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Dropout03EncoderDecoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KernelRegularizerL1Encoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KernelRegularizerL1Decoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KernelRegularizerL2Encoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KernelRegularizerL2Decoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KernelRegularizerL2EncoderDecoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ActivityRegularizerL1Encoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ActivityRegularizerL1Decoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ActivityRegularizerL1EncoderDecoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ActivityRegularizerL2Encoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ActivityRegularizerL2Decoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ActivityRegularizerL2EncoderDecoder |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
AveragePooling |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
NoPooling |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
NoDense |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
LatentDim4096 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
LatentDim2048 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|