Datasetpublic

Primusvandalus/grocery-images-5class

grocery-images-5class 1,340 photographs across five grocery classes, collected for a university machine-learning course project: an image classifier whose predictions chain into a grocery text-classification pipeline. class images apple 300 banana 260 bread_roll 260 cheese 260 tomato 260 Collection and labelling Images were gathered through the Pexels API in April 2026 and hand-labelled with a custom keyboard-driven dashboard; every… See the full description on the dataset page: https://huggingface.co/datasets/Primusvandalus/grocery-images-5class.

sourceHugging Faceotherupdated 3mo agoView on Hugging Face

0likes398downloads

Dataset Card

grocery-images-5class

1,340 photographs across five grocery classes, collected for a university machine-learning course project: an image classifier whose predictions chain into a grocery text-classification pipeline.

class	images
apple	300
banana	260
bread_roll	260
cheese	260
tomato	260

Collection and labelling

Images were gathered through the Pexels API in April 2026 and hand-labelled with a custom keyboard-driven dashboard; every accept/reject decision was logged, and a written rubric governed edge cases (the dashboard, rubric, and label log live in the companion code repository). Filenames carry the source photo id as {class}_{pexels_photo_id}.jpg, so every image traces back to its Pexels page.

Splits

splits_seed42.json is the filename-level 70/15/15 train/val/test manifest (938/201/201) used by the companion repository. Folder structure is one directory per class, so the dataset also loads directly:

python

from datasets import load_dataset
ds = load_dataset("imagefolder", data_dir=".")

License

The photographs are provided under the Pexels license. They are redistributed here for research and educational use; photo ids are preserved so any image can be checked against its source.