| title: CURE | |
| emoji: 🎛️ | |
| colorFrom: blue | |
| colorTo: indigo | |
| sdk: gradio | |
| sdk_version: 5.38.0 | |
| app_file: app.py | |
| python_version: "3.10" | |
| models: | |
| - ses7720/CURE | |
| tags: | |
| - image-restoration | |
| - controllable-image-restoration | |
| - low-light-enhancement | |
| - dehazing | |
| - deraining | |
| - desnowing | |
| preload_from_hub: | |
| - ses7720/CURE CURE_restorer.tar,OneRestore_embedder.tar | |
| # CURE: Controllable Unified Image Restoration | |
| This Gradio Space exposes the five CURE inference modes: | |
| - full one-step restoration; | |
| - ratio-controlled restoration; | |
| - selective degradation removal; | |
| - identity/no-restoration inference; and | |
| - ordered two-stage restoration. | |
| The model weights are loaded from [ses7720/CURE](https://huggingface.co/ses7720/CURE). | |
| The source code and command-line inference tools are available on | |
| [GitHub](https://github.com/bo-oseng/CURE). | |
| Each tab includes representative CCDD-11 sample images. The samples cover | |
| three different scenes, each with all 11 supported degradation combinations, | |
| making it easy to compare controls without uploading an image first. | |