This Repository is a part of Deep Learning Course by MATHWORKS. Every Musical intrument have a specific pattern in thier respective spectrograph images. Using this images, a deep neural network can be developed which can identify to which Instrument the spectrograph belongs.
The Network has been trained to detect 14 instruments
- Bells
- Cello
- Clarinet
- Crotales
- Double Bass
- Flute
- Piano
- Saxophone
- Trombone
- Trumpet
- Vibraphone
- Viola
- Violin
- Xylophone
The spectrogram images of these instruments were obtained from MATHWORKS, in association with University of Iowa Electronic Music Studios. Alex Net was modified with Transfer learning, 70.06% acccuracy was achieved. Further modifications on training options led to an accuracy of 88.62%, this was with the Transfer Learning. The Results have been shared as "result_1.png" and "result_2.png"
Neural network "Spec_New_Network based on speech recognition architecture, predicted the instrumests with an accuracy of 97.48%. the training progress can be visualized in the "result_3.png"