Hugging Face's logo

OpticAI
/
LUCID

Image-to-Image
image-restoration
low-light-enhancement
deflare
diffusion
nighttime-photography
siggraph-2026
Model card Files
xet
 Community
LUCID
File size: 2,598 Bytes
10db347
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7f6e930
10db347
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
---
license: other
pipeline_tag: image-to-image
tags:
  - image-restoration
  - low-light-enhancement
  - deflare
  - diffusion
  - nighttime-photography
  - siggraph-2026
---

# LUCID

**LUCID: Learning Unified Control for Image Deflaring and Exposure Mastery in Nighttime Photography**

LUCID is a unified framework for nighttime image restoration. It targets the coupled degradations that often appear together in night photography: underexposure, intense flare, ghosting artifacts, and visible light sources.

> Photography is the art of painting with light, yet nighttime scenes are shaped by competing degradations: intense flares obscure scene structure, while photon-limited regions collapse into noise. Conventional approaches address these factors in isolation, overlooking the fact that these degradations are fundamentally entangled. LUCID reframes nighttime restoration as a continuous and controllable process rather than a fixed correction, restoring challenging nighttime images with flexible control over exposure, light sources, flare, and ghosting artifacts.

## Checkpoints

This repository contains the pretrained checkpoints used by the LUCID codebase:

| File | Description |
| --- | --- |
| `LUCID_main/model_40000.pkl` | Main LUCID restoration checkpoint |
| `Flare_Disentangle/latest.pth` | Flare disentanglement network checkpoint |

The base diffusion model is **SD-Turbo** and should be downloaded separately:

https://huggingface.co/stabilityai/sd-turbo

## Usage

Clone the code repository:

```bash
git clone https://github.com/frakenation/LUCID.git
cd LUCID
```

Download these checkpoints:

```bash
git lfs install
git clone https://huggingface.co/Unswear/LUCID lucid_weights
```

Example inference:

```bash
python -m src.inference \
  --input_dir ./data/test/input \
  --pretrained_model_name_or_path stabilityai/sd-turbo \
  --model_path ./lucid_weights/LUCID_main/model_40000.pkl \
  --flare_disentanglement_path ./lucid_weights/Flare_Disentangle/latest.pth \
  --output_dir ./results/lucid \
  --resolution 512 \
  --timestep 199 \
  --ms_unet \
  --inference_mode cfg_guidance \
  --cfg_scale 1.05 \
  --device cuda
```

For multi-scale exposure control, use `src.inference_cfg_index` or the provided shell scripts in the GitHub repository.

## Links

- Project page: https://xiaoyunyuan.net/index.html?project=lucid
- Paper: https://arxiv.org/abs/2606.06901
- Code: https://github.com/frakenation/LUCID
- Video: https://www.youtube.com/watch?v=AGPLSiZcK_I

## Citation

If you find this project useful, please cite the paper from the project page or arXiv.