Reducing the feature divergence of RGB and near-infrared images using Switchable Normalization

Abstract

Visual pattern recognition over agricultural areas is an important application of aerial image processing. In this paper,weconsiderthemulti-modalitynatureofagricultural aerial images and show that naively combining different modalities together without taking the feature divergence intoaccountcanleadtosub-optimalresults. Thus,weapply a SwitchableNormalization blockto ourDeepLabV3+ segmentation model to alleviate the feature divergence. Using the popular symmetric Kullback–Leibler divergence measure, we show that our model can greatly reduce the divergence between RGB and near-infrared channels. Together with a hybrid loss function, our model achieves nearly 10% improvements in mean IoU over previously published baseline.