Loading...
Loading...
Dataset URI:
doi:10.7910/DVN/K2QUM4This dataset contains 21,952 RGB images of five tropical fruits (banana, durian, guava, mango and papaya) labelled for binary ripeness classification (ripe / unripe), a relative depth map for every image, and the Python code used in the accompanying study. Images were compiled from public sources: BananaID (banana); two Roboflow durian datasets; the Guava Fruit Dataset for Classification; the Kaggle Fruit Image Dataset (22 classes) and the Mango Shelf-Life dataset (mango); the Kaggle Fruit Image Dataset and Roboflow Papaya-maturity v15 (papaya). Original labels were mapped to ripe / unripe, and intermediate, defective, perished and multi-label images were removed. Images are organised in ten folders named "ripe " and "unripe ". Counts (ripe / unripe): banana 3,500 / 3,380; durian 2,277 / 1,727; guava 1,248 / 1,061; mango 2,136 / 2,507; papaya 2,057 / 2,059. Depth maps were generated with Depth Anything V2 Small (model input 364 px) and stored as 16-bit PNG files in a folder structure parallel to the images. Each map is min-max normalised per image (larger value = closer; longest side at most 512 px). They are relative depth, not metric depth. Important for users: some sources contain augmented copies (flips and rotations). Using perceptual hashing (Hamming distance at most 6), 61.7% of the images fall into near-duplicate groups. Train/test splits should be made by group rather than by individual image; otherwise accuracy is overestimated (by about 4 F1 points in our experiments). The grouping code is included. The code covers depth map generation, depth-assisted fruit segmentation, extraction of 29 colour, texture and shape features and 13 geometric depth features, and a 5-fold stratified grouped cross-validation ablation study. Source images remain under the licences of their original datasets; see the included README files.
Keyword(s):
MeSH Term(s):
Grants and Funding:
No Grants or Funding Data Available
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
No Supplementary Data Available
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)