Deep convolutional neural networks (DCNNs) have been used to achievestate-of-the-art performance on many computer vision tasks (e.g., objectrecognition, object detection, semantic segmentation) thanks to a largerepository of annotated image data. Large labeled datasets for other sensormodalities, e.g., multispectral imagery (MSI), are not available due to thelarge cost and manpower required. In this paper, we adapt state-of-the-art DCNNframeworks in computer vision for semantic segmentation for MSI imagery. Toovercome label scarcity for MSI data, we substitute real MSI for generatedsynthetic MSI in order to initialize a DCNN framework. We evaluate our networkinitialization scheme on the new RIT-18 dataset that we present in this paper.This dataset contains very-high resolution MSI collected by an unmannedaircraft system. The models initialized with synthetic imagery were less proneto over-fitting and provide a state-of-the-art baseline for future work.
展开▼
