AppleGrowthVision

Large-scale heterogeneous apple tree data for BBCH classification, Apple Detection, and 3D reconstruction of orchards with sparse data.

Project Page

BBCH classification

1. Download the data
2. Run prepare_bbch.py to copy and rename the data for bbch classification. The data can then be used via the torchvision.datasets.ImageFolder class.

python prepare_bbch.py -p <path to directory "AppleGrowthVision"> -o <path to directory of final dataset>

Stereo Reconstruction

The script calib_conversion.py provides a function convert_xml_to_cameras(path) to retrieve the calibration parameters for a given capture date. It returns the parameters in openCV format for each camera (L, R).

Dataset

The dataset consists of two subsets that have different modalities. Brandenburg was captured using a calibrated stereo setup. Saxony was captured by a smartphone. The data can be found here.

Brandenburg

- calib/
  - <capture date>.xml
- data/
  - <capture date>/
    - L_*.jpg
    - R_*.jpg
  - annotations.json

The brandenburg data has a directory calib which contains all calibration parameters per capture date. It might be converted using calib_conv.py.

The directory data contains images per capture date named as <camera>_<image pair>.jpg. Additionally, bounding boxes for apple detection are provided in the annotations.json.

Saxony

- 2024_counted_trees
  - counted_trees_images/
    - <row id>_<tree id>/
      - *.jpg
  - annotations.json
  - 2024_counted_trees.json
- <year>/
  - <date>_<time>_pillnitz_<growthstage>_tree_<row id>-<tree id>_<picture id>.jpg
- annotations.json

The saxony data contains directories for each year. The images inside the year directory are named as <date>_<time>_pillnitz_<growthstage>_<row id>-<tree id>_<picture id>.jpg.

Valid growthstages are:

• apple blossom
• apple small fruit
• apple middle fruit
• apple fruit

The directory 2024_counted_trees contains human annotations for 10 trees with 6 orbital images and annotated bounding boxes:

row-index	tree-index	tree-id	apple count
36	2	36.2	23
36	5	36.5	1
37	2	37.2	4
44	65	44.65	28
45	9	45.9	14
45	49	45.49	32
46	4	46.4	11 + 1 bad quality
46	63	46.63	24 + 2 bad quality
47	8	47.8	15
47	12	47.12	28 + 4 bad quality

Citation

If the provided data is used or any of the provided scripts, please cite:

@InProceedings{von_Hirschhausen_2025_CVPR,
    author    = {von Hirschhausen, Laura-Sophia and Magnusson, Jannes S. and Kovalenko, Mykyta and Boye, Fredrik and Rawat, Tanay and Eisert, Peter and Hilsmann, Anna and Pretzsch, Sebastian and Bosse, Sebastian},
    title     = {AppleGrowthVision: A large-scale stereo dataset for phenological analysis, fruit detection, and 3D reconstruction in apple orchards},
    booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) Workshops},
    month     = {June},
    year      = {2025},
    pages     = {5443-5450}
}