Bilinear CNN implementation using keras
Bilinear CNN is simple yet powerful model for Fine-Grained Visual Classification task.
it is still competitive though developed more than 3 years ago.
reference here
http://vis-www.cs.umass.edu/bcnn/docs/bcnn_iccv15.pdf
In this demo, I implemented B-CNN D-D model with keras.
B-CNN leverage pre-trained networks.
D-D model use two identical VGG16 as backbone.
Caltech-UCSD Birds-200-2011 (CUB-200-2011) is an image dataset with photos of 200 bird species.
It contains total 11,788 images.
reference here
http://www.vision.caltech.edu/visipedia/CUB-200-2011.html