Thanks for your interests in our musical instrument embedding work "Use of Speaker Recognition Approaches for Learning and Evaluating Embedding Representations of Musical Instrument Sounds" in IEEE/ACM Transactions on Audio, Speech, and Language Processing!! The code is continuely tidying up. The current version still has redundant part for development. Feel free to contact me（[email protected]) if you have any problems. I generally reply emails in 24 hours.
Train:
python train.py config/resnet34_sinc_nsynth.py
Inference:
python inference.py --config_file=config/resnet34_sinc_nsynth.py --pretrain_model_pth='log/nsynth_sinc_midi_aug_l8_asm_s100/8_77.h5'
EER: 0.02935