Datasets:
Link paper and add sample usage
#1
by nielsr HF Staff - opened
README.md
CHANGED
|
@@ -1,24 +1,25 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
| 2 |
license: cc-by-nc-4.0
|
|
|
|
|
|
|
| 3 |
task_categories:
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
language:
|
| 7 |
-
- en
|
| 8 |
tags:
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
- depth-estimation
|
| 15 |
-
size_categories:
|
| 16 |
-
- 1K<n<10K
|
| 17 |
---
|
| 18 |
|
| 19 |
# D-Synth: Synthetic Dermoscopic Dataset with Pixel-Perfect 3D Information
|
| 20 |
|
| 21 |
-
D-Synth is the first synthetic dermoscopic dataset providing pixel-perfect 3D ground truth (metric depth, surface normals, camera intrinsics) for monocular depth estimation in dermatology.
|
|
|
|
|
|
|
| 22 |
|
| 23 |
## Overview
|
| 24 |
|
|
@@ -30,6 +31,26 @@ D-Synth is the first synthetic dermoscopic dataset providing pixel-perfect 3D gr
|
|
| 30 |
- `generation_params.json` — full rendering parameters
|
| 31 |
- (subset) `render_rgb.png`, `render_depth.png`, `render_meta.json` — additional render variants
|
| 32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
## Generation pipeline
|
| 34 |
|
| 35 |
D-Synth extends [S-SYNTH](https://github.com/DIDSR/ssynth-release) (Kim et al., MICCAI 2024) with:
|
|
@@ -72,4 +93,4 @@ CC BY-NC 4.0 (research / non-commercial use). For other uses, please contact the
|
|
| 72 |
|
| 73 |
- **Code & training scripts**: https://github.com/hectorcarrion/dermdepth
|
| 74 |
- **Fine-tuned DermDepth checkpoints**: https://huggingface.co/hcarrion/DermDepth
|
| 75 |
-
- **Base depth model**: https://huggingface.co/Ruicheng/moge-2-vitl-normal
|
|
|
|
| 1 |
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
license: cc-by-nc-4.0
|
| 5 |
+
size_categories:
|
| 6 |
+
- 1K<n<10K
|
| 7 |
task_categories:
|
| 8 |
+
- depth-estimation
|
| 9 |
+
- image-to-image
|
|
|
|
|
|
|
| 10 |
tags:
|
| 11 |
+
- dermatology
|
| 12 |
+
- medical-imaging
|
| 13 |
+
- 3d-reconstruction
|
| 14 |
+
- synthetic-data
|
| 15 |
+
- skin-lesion
|
|
|
|
|
|
|
|
|
|
| 16 |
---
|
| 17 |
|
| 18 |
# D-Synth: Synthetic Dermoscopic Dataset with Pixel-Perfect 3D Information
|
| 19 |
|
| 20 |
+
D-Synth is the first synthetic dermoscopic dataset providing pixel-perfect 3D ground truth (metric depth, surface normals, camera intrinsics) for monocular depth estimation in dermatology. It was introduced in [DermDepth: Toward Monocular Metric Scale 3D Reconstruction Models for Dermatology](https://huggingface.co/papers/2607.13010) (Carrión & Norouzi, MICCAI 2026).
|
| 21 |
+
|
| 22 |
+
[Project Page / GitHub](https://github.com/hectorcarrion/dermdepth) | [Paper](https://huggingface.co/papers/2607.13010)
|
| 23 |
|
| 24 |
## Overview
|
| 25 |
|
|
|
|
| 31 |
- `generation_params.json` — full rendering parameters
|
| 32 |
- (subset) `render_rgb.png`, `render_depth.png`, `render_meta.json` — additional render variants
|
| 33 |
|
| 34 |
+
## Sample Usage
|
| 35 |
+
|
| 36 |
+
You can download the dataset locally using the Hugging Face CLI:
|
| 37 |
+
|
| 38 |
+
```bash
|
| 39 |
+
hf download hcarrion/D-Synth --repo-type dataset --local-dir data/dermdepth_train/dsynth
|
| 40 |
+
```
|
| 41 |
+
|
| 42 |
+
Or programmatically via Python:
|
| 43 |
+
|
| 44 |
+
```python
|
| 45 |
+
from huggingface_hub import snapshot_download
|
| 46 |
+
|
| 47 |
+
snapshot_download(
|
| 48 |
+
repo_id='hcarrion/D-Synth',
|
| 49 |
+
repo_type='dataset',
|
| 50 |
+
local_dir='data/dermdepth_train/dsynth'
|
| 51 |
+
)
|
| 52 |
+
```
|
| 53 |
+
|
| 54 |
## Generation pipeline
|
| 55 |
|
| 56 |
D-Synth extends [S-SYNTH](https://github.com/DIDSR/ssynth-release) (Kim et al., MICCAI 2024) with:
|
|
|
|
| 93 |
|
| 94 |
- **Code & training scripts**: https://github.com/hectorcarrion/dermdepth
|
| 95 |
- **Fine-tuned DermDepth checkpoints**: https://huggingface.co/hcarrion/DermDepth
|
| 96 |
+
- **Base depth model**: https://huggingface.co/Ruicheng/moge-2-vitl-normal
|