Automated Cloud Removal on High-Altitude UAV Imagery Through Deep Learning on Synthetic Data
Ryan A Wells · OakTrust (Texas A&M University Libraries) · 2019
New theories and applications of deep learning have been discovered and implemented within the field of machine learning recently. The high degree of effectiveness of deep learning models span across many domains including image processing and enhancement. Specifically, the automated removal of clouds, smoke, and haze from images has become a prominent and pertinent field of research. In this paper, I propose an analysis and synthetic training data variant for the All-in-One Dehazing Network (AOD-Net) architecture that performs better on removing clouds and haze; most specifically on high altitude unmanned aerial vehicles (UAVs) images.