Spaces:
Running
Running
title: CCSR Upscaler | |
emoji: π | |
colorFrom: gray | |
colorTo: indigo | |
sdk: gradio | |
sdk_version: 4.42.0 | |
app_file: app.py | |
pinned: true | |
short_description: Upscale an image using CCSR | |
# CCSR Upscaler | |
This Gradio space implements the Continuous Contrastive Super-Resolution (CCSR) model for image upscaling. CCSR is a state-of-the-art super-resolution method that can upscale images while preserving details and enhancing quality. | |
## Features | |
- Upload any image for upscaling | |
- Adjust super-resolution scale (1x to 8x) | |
- Fine-tune parameters like t_max and t_min | |
- Choose from different color fixing methods | |
## How to Use | |
1. Upload an image you want to upscale | |
2. Adjust the SR Scale slider to set the upscaling factor | |
3. Fine-tune t_max and t_min values if desired | |
4. Select a color fixing method from the dropdown | |
5. Click "Submit" to generate the upscaled image | |
## Model Details | |
This space uses the CCSR model trained on real-world images. The model checkpoint and configuration are loaded from: | |
- Checkpoint: `weights/real-world_ccsr.ckpt` | |
- Config: `configs/model/ccsr_stage2.yaml` | |
## Requirements | |
The main dependencies for this project are listed in the `requirements.txt` file, including: | |
- torch | |
- torchvision | |
- gradio | |
- einops | |
- pytorch-lightning | |
- omegaconf | |
- open-clip-torch | |
- xformers | |
- taming-transformers | |
## Acknowledgements | |
This implementation is based on the CCSR model. For more details about the original work, please refer to the [CCSR GitHub repository](https://github.com/camenduru/CCSR). |