NTIRE 2020 tested how image super-resolution methods perform when paired low- and high-resolution training images are unavailable. Its two tracks highlight the role of degradation modeling, unpaired data, and human evaluation in real-world image enhancement.