Datasets:
Card: size figures in decimal GB, measured from the file listing
Browse files
README.md
CHANGED
|
@@ -8,7 +8,7 @@ size_categories: [n<1K]
|
|
| 8 |
|
| 9 |
# Single Tree — High-Density Photogrammetry Dataset
|
| 10 |
|
| 11 |
-
**812 photos of one mature deciduous tree, flown from 0.5 to 7.9 m above the ground.
|
| 12 |
807 align, and the solved camera poses ship with it in **COLMAP format** — point a Gaussian
|
| 13 |
splatting pipeline straight at it, no structure-from-motion run required.** CC BY 4.0.
|
| 14 |
|
|
@@ -16,7 +16,7 @@ splatting pipeline straight at it, no structure-from-motion run required.** CC B
|
|
| 16 |
|
| 17 |
### **[→ huggingface.co/datasets/Matt1up/tree-minnetonka-photogrammetry](https://huggingface.co/datasets/Matt1up/tree-minnetonka-photogrammetry)**
|
| 18 |
|
| 19 |
-
Browse the **Files** tab and take what you want — no account needed. The
|
| 20 |
there because GitHub won't host files that size; this repo holds docs, checksums and poses.
|
| 21 |
|
| 22 |
[Sample image](https://huggingface.co/datasets/Matt1up/tree-minnetonka-photogrammetry/resolve/main/images/The_Tree-1.jpg)
|
|
@@ -73,7 +73,7 @@ Full project write-up: **[mattguertin.com/portfolio/tree](https://mattguertin.co
|
|
| 73 |
|
| 74 |
| | |
|
| 75 |
|---|---|
|
| 76 |
-
| **Images** | 812 JPEG ·
|
| 77 |
| **Alignment** | 807 / 812 images solve |
|
| 78 |
| **Sensor** | Hasselblad L1D-20c — 1" 20 MP CMOS (DJI Mavic 2 Pro) |
|
| 79 |
| **Resolution** | 5464 × 3640 |
|
|
@@ -123,7 +123,7 @@ exactly what those pipelines ingest.
|
|
| 123 |
| `poses/cameras.csv` | the same data as one table |
|
| 124 |
| `poses/suspect_cameras.txt` | 4 cameras with implausible solved focal lengths |
|
| 125 |
| `poses/unaligned.txt` | the 5 that did not solve |
|
| 126 |
-
| `tiepoints.ply` | 1.2M sparse points,
|
| 127 |
|
| 128 |
Solved in RealityScan 2.2, which is free, then converted to COLMAP format by
|
| 129 |
[`scripts/xmp-to-colmap.py`](https://github.com/Matt1Up/tree-photogrammetry-dataset/blob/main/scripts/xmp-to-colmap.py). The converter does not assume the
|
|
@@ -131,7 +131,7 @@ camera-frame convention — it tests both by reprojecting tie points and keeps w
|
|
| 131 |
them in front of the cameras. See [`poses/colmap/README.md`](https://github.com/Matt1Up/tree-photogrammetry-dataset/blob/main/poses/colmap/README.md).
|
| 132 |
|
| 133 |
```bash
|
| 134 |
-
hf download Matt1up/tree-minnetonka-photogrammetry --repo-type dataset --local-dir ./tree --include 'colmap/*' # points3D.txt,
|
| 135 |
```
|
| 136 |
|
| 137 |
The 5 that didn't align are all from the low and mid tiers, none from `The_Tree`.
|
|
@@ -159,7 +159,7 @@ pip install -U huggingface_hub
|
|
| 159 |
hf download Matt1up/tree-minnetonka-photogrammetry --repo-type dataset --local-dir ./tree
|
| 160 |
```
|
| 161 |
|
| 162 |
-
Take part of it with `--include`: `'sample/*'` (~
|
| 163 |
`'images/*'`, `'colmap/*'`, or `'images/Original_low*'` for one capture group.
|
| 164 |
|
| 165 |
**Everything, as a git repo** (needs git-lfs):
|
|
@@ -175,7 +175,7 @@ SHA-256 list:
|
|
| 175 |
```bash
|
| 176 |
git clone https://github.com/Matt1Up/tree-photogrammetry-dataset && cd tree-photogrammetry-dataset
|
| 177 |
|
| 178 |
-
./scripts/download.sh --sample # ~
|
| 179 |
./scripts/download.sh --full # everything
|
| 180 |
./scripts/download.sh --colmap # points3D.txt for splatting
|
| 181 |
./scripts/download.sh --group Original_low --group Original_mid
|
|
|
|
| 8 |
|
| 9 |
# Single Tree — High-Density Photogrammetry Dataset
|
| 10 |
|
| 11 |
+
**812 photos of one mature deciduous tree, flown from 0.5 to 7.9 m above the ground. 15.1 GB.
|
| 12 |
807 align, and the solved camera poses ship with it in **COLMAP format** — point a Gaussian
|
| 13 |
splatting pipeline straight at it, no structure-from-motion run required.** CC BY 4.0.
|
| 14 |
|
|
|
|
| 16 |
|
| 17 |
### **[→ huggingface.co/datasets/Matt1up/tree-minnetonka-photogrammetry](https://huggingface.co/datasets/Matt1up/tree-minnetonka-photogrammetry)**
|
| 18 |
|
| 19 |
+
Browse the **Files** tab and take what you want — no account needed. The 15 GB of imagery lives
|
| 20 |
there because GitHub won't host files that size; this repo holds docs, checksums and poses.
|
| 21 |
|
| 22 |
[Sample image](https://huggingface.co/datasets/Matt1up/tree-minnetonka-photogrammetry/resolve/main/images/The_Tree-1.jpg)
|
|
|
|
| 73 |
|
| 74 |
| | |
|
| 75 |
|---|---|
|
| 76 |
+
| **Images** | 812 JPEG · 15.10 GB |
|
| 77 |
| **Alignment** | 807 / 812 images solve |
|
| 78 |
| **Sensor** | Hasselblad L1D-20c — 1" 20 MP CMOS (DJI Mavic 2 Pro) |
|
| 79 |
| **Resolution** | 5464 × 3640 |
|
|
|
|
| 123 |
| `poses/cameras.csv` | the same data as one table |
|
| 124 |
| `poses/suspect_cameras.txt` | 4 cameras with implausible solved focal lengths |
|
| 125 |
| `poses/unaligned.txt` | the 5 that did not solve |
|
| 126 |
+
| `tiepoints.ply` | 1.2M sparse points, 80 MB — hosted with the images, not in git |
|
| 127 |
|
| 128 |
Solved in RealityScan 2.2, which is free, then converted to COLMAP format by
|
| 129 |
[`scripts/xmp-to-colmap.py`](https://github.com/Matt1Up/tree-photogrammetry-dataset/blob/main/scripts/xmp-to-colmap.py). The converter does not assume the
|
|
|
|
| 131 |
them in front of the cameras. See [`poses/colmap/README.md`](https://github.com/Matt1Up/tree-photogrammetry-dataset/blob/main/poses/colmap/README.md).
|
| 132 |
|
| 133 |
```bash
|
| 134 |
+
hf download Matt1up/tree-minnetonka-photogrammetry --repo-type dataset --local-dir ./tree --include 'colmap/*' # points3D.txt, 58 MB
|
| 135 |
```
|
| 136 |
|
| 137 |
The 5 that didn't align are all from the low and mid tiers, none from `The_Tree`.
|
|
|
|
| 159 |
hf download Matt1up/tree-minnetonka-photogrammetry --repo-type dataset --local-dir ./tree
|
| 160 |
```
|
| 161 |
|
| 162 |
+
Take part of it with `--include`: `'sample/*'` (~440 MB, look before committing to 16 GB),
|
| 163 |
`'images/*'`, `'colmap/*'`, or `'images/Original_low*'` for one capture group.
|
| 164 |
|
| 165 |
**Everything, as a git repo** (needs git-lfs):
|
|
|
|
| 175 |
```bash
|
| 176 |
git clone https://github.com/Matt1Up/tree-photogrammetry-dataset && cd tree-photogrammetry-dataset
|
| 177 |
|
| 178 |
+
./scripts/download.sh --sample # ~440 MB, look before committing to 16 GB
|
| 179 |
./scripts/download.sh --full # everything
|
| 180 |
./scripts/download.sh --colmap # points3D.txt for splatting
|
| 181 |
./scripts/download.sh --group Original_low --group Original_mid
|