Remote Sensing Object Counting Through Regression Ensembles and Learning to Rank

Remote sensing object counting (RSOC) is finding applications in many fields. Global regression is a long-ignored method for object counting, though it needs much less manual annotations than the alternatives. This work revisits global regression and improves it in two ways-one way is by replacing one single regressor with a deep ensemble, and the other is by breaking down global regression into two easier and smaller problems: learning to rank (L2R) and linear transformation (LT). To this end, we make a probably approximately correct (PAC)-Bayesian analysis of regression ensembles and give an upper bound for their generalization error, offering new theoretical insight into ensemble learning. We also adapt a ranking metric optimization scheme to suit object counting, elegantly handling the L2R problem with gradient descent. Furthermore, based on our theoretical perspective, we provide a novel way of building deep regression ensembles, on which the ambiguity constraint is imposed. Then, by incorporating L2R into a deep ensemble, we propose a new counting model called the ensemble of first-rank-then-estimate networks (eFreeNet). Our extensive evaluation on six benchmarks shows that the eFreeNet exhibits compelling performance across the board while being more annotation-efficient than other methods. Our source code is publicly available at https://github.com/huangyongbobo/eFreeNet.

Automated summary

Remote sensing object counting (RSOC) has various applications, and this study improves the historical method of global regression by replacing a single regressor with a deep ensemble and breaking down the problem into learning to rank (L2R) and linear transformation (LT). The study offers new theoretical insights into ensemble learning and provides a novel way of building deep regression ensembles. The proposed counting model, eFreeNet, exhibits superior performance and is more annotation-efficient than other methods on six benchmarks.