for multi-label image classification, if train the net with single label data. for example,i have three class[dog, cat, bird],i only feed the net single label data ,like dog, bird, cat. finally ,i use a image with tow class ,i.e [dog,cat].the net will give the correct result? or have to train with mulit-label images ?
for multi-label image classification, if train the net with single label data. for example,i have three class[dog, cat, bird],i only feed the net single label data ,like dog, bird, cat. finally ,i use a image with tow class ,i.e [dog,cat].the net will give the correct result? or have to train with mulit-label images ?