6 benchmark datasets and 13 model architectures for image super-resolution. PyTorch implementations for architectures such as EDSR and RCAN are provided alongside standard evaluation sets like Set5 and Set14.
Use Cases
- Train a super-resolution neural network using the DIV2K training images and provided scripts.
- Benchmark image upscaling quality using PSNR and SSIM metrics on the Set14 dataset.
- Perform image restoration using pre-trained weights for the RCAN model architecture.
Strengths
- Includes the DIV2K dataset featuring 800 high-resolution images for training.
- Provides pre-trained weights in PyTorch .pth format for models including EDSR and RCAN.
- Supports evaluation on 6 standard benchmarks including Set5, Set14, BSD100, and Urban100.