source code *.ipynb
pre-trained model ./model
base_line source code ./baseline
*net.ipynb the netowrok architecture of model
*loss.ipynb the loss curves of model
*dataloader.ipynb the dataloader of model
*testAccuracy.ipynb the accuracy of model conducted on 3-piece fragments
Different_SongParts_Test.ipynb the accuracy of model conducted on multi-piece fragements
Experiment_Visualization.ipynb exact features and 2d-visualization
we have presented a novel triplet Siamese network dubbed as triple-wise similarity with recurrent neural network (TSN-R) for learning sequential patterns in a self- supervised way.
we reveal the reasons for the good performance of TSN-R network through the 2d-visualization of features.