FINE-BITE IMAGE CLASSIFICATION BY EXPLORING LABELS FROM A BIPARTITE GRAPH

摘要

Disclosed are systems and methods for deep learning and classifying images of objects by receiving images of objects for training or classifying the objects; for producing fine-grained labels of the objects; for providing object images to a multiple class convolutional neural network (CNN) having a softmax layer and a final fully connected layer to explicitly model bi-directional graph (BGLs) labels; and to optimize the CNN with global backward propagation.