Pre-trained representations are becoming crucial for many NLP and percep...
Deep image embedding provides a way to measure the semantic similarity o...
We present Open Images V4, a dataset of 9.2M images with unified annotat...
Most deep architectures for image classification--even those that are tr...
Current approaches for fine-grained recognition do the following: First,...