Pinned Repositories
EE488_AI_Convergence_Capstone_Design_Anomaly_Detection_2022spring
AutoSusPed
Research in automatic piano music generation has only recently started to involve piano pedals as a part of the task. In this work, we train various neural network architectures with piano sustain pedal control-change (CC) data using different categories within the MAESTRO classical piano music dataset to study the performances of basic models and test the suitability of neural networks in an automatic piano pedal styling task. By changing the temporal scanning range of convolution kernels and the depth of the network structure, we show that both factors are relevant in the accuracy of pedaling style prediction. Currently, our best CNN-based Auto-SusPed model predicts a specific composer’s pedaling style and a specific musical era’s style with accuracies of around 90%.
ic_ib_nd
inter-channel information-based noise detector
jim8220
Config files for my GitHub profile.
WS-AUROC
official code for 'Performance metric for multiple anomaly score distributions having discrete severity levels'
jim8220's Repositories
jim8220/AutoSusPed
Research in automatic piano music generation has only recently started to involve piano pedals as a part of the task. In this work, we train various neural network architectures with piano sustain pedal control-change (CC) data using different categories within the MAESTRO classical piano music dataset to study the performances of basic models and test the suitability of neural networks in an automatic piano pedal styling task. By changing the temporal scanning range of convolution kernels and the depth of the network structure, we show that both factors are relevant in the accuracy of pedaling style prediction. Currently, our best CNN-based Auto-SusPed model predicts a specific composer’s pedaling style and a specific musical era’s style with accuracies of around 90%.
jim8220/ic_ib_nd
inter-channel information-based noise detector
jim8220/jim8220
Config files for my GitHub profile.
jim8220/WS-AUROC
official code for 'Performance metric for multiple anomaly score distributions having discrete severity levels'