traiNNer-redux
Deep learning training framework for image super resolution and restoration.
APACHE-2.0 License
Stars
23
Pretrained models which serve as a showcase of traiNNer-redux. These models were trained on a single RTX 4090 and surpass the metrics of paper models trained on 4 A100 GPUs. These models were trained following the methods outlined in the training guidelines.
DAT2
| Model | Urban100 PSNR | Urban100 SSIM |
|---|---|---|
| DAT-2 x2 | 34.31 | 0.9457 |
| 2x_DF2K_Redux_DAT2 | 34.33 | 0.9471 |
| DAT-2 x4 | 27.86 | 0.8341 |
| 4x_DF2K_Redux_DAT2 | 27.86 | 0.8372 |
PLKSR
| Model | Urban100 PSNR | Urban100 SSIM |
|---|---|---|
| PLKSR DF2K x2 | 33.36 | 0.9395 |
| 2x_DF2K_Redux_RealPLKSRLayerNorm | 33.44 | 0.9412 |
| PLKSR DF2K x4 | 26.85 | 0.8097 |
| 4x_DF2K_Redux_RealPLKSRLayerNorm | 26.94 | 0.8140 |