README.md

---
license: cc-by-4.0
task_categories:
  - image-to-image
tags:
  - medical-imaging
  - surgical-smoke
  - desmoking
  - laparoscopy
pretty_name: SeeThroughSmoke
size_categories:
  - 1K<n<10K
configs:
  - config_name: default
    data_files:
      - split: train
        path: "train/metadata.csv"
      - split: test
        path: "test/metadata.csv"
---

# SeeThroughSmoke Dataset

Paired surgical smoke / ground-truth image dataset for training and evaluating smoke removal (desmoking) models in laparoscopic surgery.

**Paper:** [Seeing Through Smoke: Surgical Desmoking for Improved Visual Perception](https://arxiv.org/abs/2603.25867) (arXiv:2603.25867)

**Project Page:** [smoke.surgeryvision.org](https://smoke.surgeryvision.org)

## Demo Videos

<table>
<tr>
<td width="50%"><b>Liver — Ours (Base Model)</b><br>
<video controls width="100%">
  <source src="https://huggingface.co/datasets/artJiang20/SeeThroughSmoke/resolve/main/videos/liver_ours_web.mp4" type="video/mp4">
</video></td>
<td width="50%"><b>Liver — Ours (Fine-tuned)</b><br>
<video controls width="100%">
  <source src="https://huggingface.co/datasets/artJiang20/SeeThroughSmoke/resolve/main/videos/liver_ours_finetune_web.mp4" type="video/mp4">
</video></td>
</tr>
<tr>
<td width="50%"><b>Heart — Ours (Base Model)</b><br>
<video controls width="100%">
  <source src="https://huggingface.co/datasets/artJiang20/SeeThroughSmoke/resolve/main/videos/heart_ours_web.mp4" type="video/mp4">
</video></td>
<td width="50%"><b>Heart — Ours (Fine-tuned)</b><br>
<video controls width="100%">
  <source src="https://huggingface.co/datasets/artJiang20/SeeThroughSmoke/resolve/main/videos/heart_outs_finetune_web.mp4" type="video/mp4">
</video></td>
</tr>
</table>

**Supplementary Video**

<video controls width="100%">
  <source src="https://huggingface.co/datasets/artJiang20/SeeThroughSmoke/resolve/main/videos/sup_video.mp4" type="video/mp4">
</video>

## Dataset Structure

Each row in the viewer shows: **Smoked Input (`image`) → Ground Truth (`gt`)** for the given `organ`.

```
SeeThroughSmoke/
├── train/                    # 3,097 paired images
│   ├── metadata.csv
│   ├── liver_kidney/         # 1,282 pairs
│   │   ├── GT/
│   │   └── smoke/
│   ├── small_Intestine/      # 648 pairs
│   ├── large_intestine/      # 596 pairs
│   ├── unknown/              # 357 pairs
│   └── uterus/               # 214 pairs
└── test/                     # 2,720 paired images
    ├── metadata.csv
    ├── brain/                # 490 pairs
    ├── heart/                # 672 pairs
    └── stomach/              # 1,558 pairs
```

## Statistics

| Split | Organ | Pairs |
|-------|-------|-------|
| Train | liver_kidney | 1,282 |
| Train | small_Intestine | 648 |
| Train | large_intestine | 596 |
| Train | unknown | 357 |
| Train | uterus | 214 |
| **Train Total** | | **3,097** |
| Test | brain | 490 |
| Test | heart | 672 |
| Test | stomach | 1,558 |
| **Test Total** | | **2,720** |

## Image Format

- Resolution: 1280 × 1024 pixels
- Format: PNG, 8-bit RGB
- Each `smoke/` image has a corresponding `GT/` image with the same filename

## Naming Convention

Files follow the pattern: `{organ}_{instance}_{side}_{frame}.png`
- `organ`: anatomical region
- `instance`: patient/scan identifier
- `side`: `left` or `right` (camera orientation)
- `frame`: frame index within the sequence

## Usage

```python
from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="artJiang20/SeeThroughSmoke",
    repo_type="dataset",
    local_dir="./SeeThroughSmoke"
)
```

Or use the `hf` CLI:

```bash
hf download artJiang20/SeeThroughSmoke --repo-type dataset
```