Real-ESRGAN is a method for blind super-resolution that uses pure synthetic data to train a model capable of enhancing details and removing artifacts from real-world images. The authors extend the ESRGAN model to Real-ESRGAN by introducing a high-order degradation modeling process to better simulate complex real-world degradations, such as camera blur, sensor noise, sharpening artifacts, and JPEG compression. They also incorporate sinc filters to model common ringing and overshoot artifacts. To improve the discriminator's capability and stabilize training dynamics, they employ a U-Net discriminator with spectral normalization. Extensive experiments on various real datasets show that Real-ESRGAN outperforms previous methods in terms of visual performance, making it more practical for real-world applications. The method is efficient and effective, achieving a good balance between local detail enhancement and artifact suppression. However, it still has limitations, such as aliasing issues and the inability to handle out-of-distribution degradations.Real-ESRGAN is a method for blind super-resolution that uses pure synthetic data to train a model capable of enhancing details and removing artifacts from real-world images. The authors extend the ESRGAN model to Real-ESRGAN by introducing a high-order degradation modeling process to better simulate complex real-world degradations, such as camera blur, sensor noise, sharpening artifacts, and JPEG compression. They also incorporate sinc filters to model common ringing and overshoot artifacts. To improve the discriminator's capability and stabilize training dynamics, they employ a U-Net discriminator with spectral normalization. Extensive experiments on various real datasets show that Real-ESRGAN outperforms previous methods in terms of visual performance, making it more practical for real-world applications. The method is efficient and effective, achieving a good balance between local detail enhancement and artifact suppression. However, it still has limitations, such as aliasing issues and the inability to handle out-of-distribution degradations.