Graphene: A Lightweight Transformer with Ising Mean-Field Refinement for Image Super-Resolution
Paper β’ 2611.11139 β’ Published
Official PyTorch checkpoints for Graphene: A Lightweight Transformer with Ising Mean-Field Refinement for Image Super-Resolution. Each file matches the corresponding checkpoint in the GitHub release.
| Task | Γ2 | Γ3 | Γ4 |
|---|---|---|---|
| Classical super-resolution | Γ2 | Γ3 | Γ4 |
| Lightweight super-resolution | Γ2 | Γ3 | Γ4 |
| Real-world super-resolution | β | β | Γ4 |
Follow the installation instructions and run these commands from the Graphene repository. The first command uses the Hugging Face CLI to download the weights.
hf download skh0830/Graphene graphene-classical_x4.pth --local-dir checkpoints
python -m tools.infer \
--config configs/graphene-classical_x4.json \
--checkpoint checkpoints/graphene-classical_x4.pth \
--input /path/to/LR --output outputs/graphene-classical_x4
For other variants, use the configuration with the same filename stem as the checkpoint. The inference guide includes the environment for real-world Γ4. See the project license and upstream notices.