Do We Need More Training Data or Better Models for Object Detection?

Xiangxin Zhu, Carl Vondrick, Deva Ramanan, Charless C. Fowlkes