Small Object Detection

In this work, we present a simple solution for increasing small object detection performance. Through a series of empirical experiments we analyze the effect of aggressive down-scaling of feature maps in a convolutional backbone.