Learning to align and match videos with kernelized temporal layers - View it on GitHub