Supplementary data for RivQNet: Deep Learning based river discharge estimation using close-range water surface imagery
The supplementary data for the "RivQNet: Deep Learning based river discharge estimation using close-range water surface imagery", recorded footage and measured ADCP data used for validation. In this study, we present RivQNet, a novel image-based method for measuring streamflow that utilizes artificial intelligence techniques and does not require subjective user input. RivQNet processes close-range, non-contact images of the water surface using a convolutional neural network architecture called FlowNet. The accuracy of RivQNet is validated through comparison with standard measurement methods and conventional optical flow methodologies. The results show that RivQNet produces accurate and dense spatial distributions of surface velocities. Streamflow data is an essential input for many hydrological and hydraulic research, modeling, and design studies, but current image-based surface velocimetry techniques that use correlation approaches can be biased if the operator is not experienced. Our goal in this study was to develop a more accurate and efficient river velocimetry method that does not rely on subjective user input.