guitarsai / basicsmusicalinstrumclassifi Goto Github PK

Hello Renato!

I have an issue and I would thanks a lot any help. I am bujilding a classifier using SVM and your notebook helped me a lot. But I could not understand why any other single predictions did not work. For example, look at the code above that i attached to the final of your SVM notebook:

x, sr = librosa.load("audio/london_phill_dataset_multi/cello/cello_A2_025_forte_arco-normal.mp3")
cello=np.array(get_features(x,sr)).reshape(1,-1)
print("cello:",svclassifier.predict(cello))

x, sr = librosa.load("audio/london_phill_dataset_multi/viola/viola_A3_025_fortissimo_arco-normal.mp3")
viola=np.array(get_features(x,sr)).reshape(1,-1)
print("viola:",svclassifier.predict(viola))

x, sr = librosa.load("audio/london_phill_dataset_multi/sax/saxophone_A3_025_forte_normal.mp3")
sax=np.array(get_features(x,sr)).reshape(1,-1)
print("sax:",svclassifier.predict(sax))

I loaded three audio files from three different instruments, but they are returning the same result:

executed in 710ms, finished 15:01:06 2019-07-02
cello: [2]
viola: [2]
sax: [2]

Can you help me undestand what is wrong in this call?

can u pls share all the perl files u have used in the code:
inst_labels
instruments_labels
mfcc_feature_vectors
max_length
mfcc_feature_vectors