DICENet: Fine-Grained Recognition via Dilated Iterative Contextual Encoding | IEEE Conference Publication | IEEE Xplore