|
|
--- |
|
|
license: cc-by-4.0 |
|
|
task_categories: |
|
|
- image-segmentation |
|
|
- keypoint-detection |
|
|
language: |
|
|
- en |
|
|
tags: |
|
|
- biology |
|
|
- agriculture |
|
|
- computer-vision |
|
|
- viticulture |
|
|
size_categories: |
|
|
- 1K<n<10K |
|
|
pretty_name: Vineyard Vision Dataset |
|
|
--- |
|
|
|
|
|
# ViViD-5k Dataset |
|
|
|
|
|
ViViD-5k is a large-scale vineyard image dataset for grape cluster analysis. It includes: |
|
|
- 5,000 images across 13 grape varieties |
|
|
- 648,000+ annotated berry centroid keypoints |
|
|
- 18,000+ grape cluster instance masks & bboxes |
|
|
|
|
|
|
|
|
## Structure |
|
|
|
|
|
```bash |
|
|
data/ |
|
|
├── imgs/ # Raw image files (JPG, PNG) |
|
|
│ ├── <img_id>.suffix |
|
|
├── anns/ # Annotation files |
|
|
│ ├── instances_updated.json # COCO-format instance segmentation masks and bboxes |
|
|
│ └── points/ # Berry keypoint annotations |
|
|
│ └── <img_id>.npy # NumPy files containing berry centroid coordinates |
|
|
└── train.txt # 4,000 training image filenames |
|
|
└── val.txt # 500 validation image filenames |
|
|
└── test.txt # 500 test image filenames |
|
|
``` |
