Sweet Pepper Dataset

The Sweet Pepper dataset is available in the following categories :

BUP19:

A sweet pepper dataset that used Gimp to annotate the images. This includes a training testing and validation set and if used should quote the following citation.

@inproceedings{halstead2020fruit,
  title={Fruit detection in the wild: The impact of varying conditions and cultivar},
  author={Halstead, Michael and Denman, Simon and Fookes, Clinton and McCool, Chris},
  booktitle={2020 Digital Image Computing: Techniques and Applications (DICTA)},
  pages={1--8},
  year={2020},
  organization={IEEE}
}

BUP19_full:

Is the full dataset color and depth images for the above BUP19 dataset. No annotations are included in this set. Please use the same citation as the BUP19 dataset.

BUP19_coco:

For future release, this will include annotations in the coco format. This is currently under construction. If you would like to know the status of this dataset please contact Dr. Michael Halstead (michael.halstead@uni-bonn.de).

BUP20:

A sweet pepper dataset which was captured in the same glasshouse as BUP19 but using PathoBot. Also, includes different cultivar, for detection and dataset use please cite:

@inproceedings{smitt2021pathobot,
  title={Pathobot: A robot for glasshouse crop phenotyping and intervention},
  author={Smitt, Claus and Halstead, Michael and Zaenker, Tobias and Bennewitz, Maren and McCool, Chris},
  booktitle={2021 IEEE International Conference on Robotics and Automation (ICRA)},
  pages={2324--2330},
  year={2021},
  organization={IEEE}
}

Similarly, for BUP20, for tracking and yield results please cite;

@article{halstead2021crop,
  title={Crop agnostic monitoring driven by deep learning},
  author={Halstead, Michael and Ahmadi, Alireza and Smitt, Claus and Schmittmann, Oliver and McCool, Chris},
  journal={Frontiers in plant science},
  volume={12},
  year={2021},
  publisher={Frontiers Media SA}
}

BUP20 Panoptic Segmentation Predictions

A dataset conteining panoptic segmentation predictions for 17990 images in the BUP20 datasets. These predictions are the output of a Mask2Former model fully supervised on BUP20’s train set. For each image instance and semantic segmentation masks are provided, as well as prediction confidence.
For panoptic segmentation predictions use please cite:

@article{smitt2023pag,
    title={PAg-NeRF: Towards fast and efficient end-to-end panoptic 3D representations for agricultural robotics},
    author={Smitt Claus and Halstead Michael and Zimmer Patrick and Laebe Thomas and Guclu Esra and Stachniss Cyrill and McCool Chris},
    journal={arXiv preprint arXiv:2309.05339},
    year={2023}
}

For further information about the BUP datasets please contact Dr. Michael Halstead (michael.halstead@uni-bonn.de)