Datasets:

manuelknott
/

banana-defects-panoptic

pretty_name: Panoptic Banana Defect Segmentatio
license: cc-by-nc-4.0
paperswithcode_id: my-dataset
bibliographic_citation: |
  @inproceedings{knott2025weakly,
    title={Weakly Supervised Panoptic Segmentation for Defect-Based Grading of Fresh Produce},
    author={Knott, Manuel and Odion, Divinefavour and Sontakke, Sameer and Karwa, Anup and Defraeye, Thijs},
    booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
    year={2025},
    pages={5471-5480}
  }
language:
  - en
tags:
  - food
  - banana
  - panoptic
  - segmentation
size_categories:
  - n<1K

My Dataset

Dataset using in the paper "Weakly Supervised Panoptic Segmentation for Defect-Based Grading of Fresh Produce".

Please note that this is the raw data provided in COCO format. Please refer to the paper's GitHub project for useful dataloader implementations. The dataloader code contains features such as train/eval split, preprocessing (merging) overlapping defects, and aggregating defect classes.