Dataset Viewer
Auto-converted to Parquet Duplicate
image
image
label
string
species
string
roast_level
string
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
Arabika_Dark
Arabika
Dark
End of preview. Expand in Data Studio

Indonesian Four-Species Coffee Bean Raw RGB Dataset (2,000 Images)

Balanced raw/standard RGB single-bean image dataset (N = 2,000 images across 16 composite classes: 4 coffee species Γ— 4 roast levels, 125 images per class) curated under the LPPM Universitas Trunojoyo Madura (UTM) Research Assistance Program in collaboration with CanopΓ©e Coffee Roastery (Mojokerto, East Java, Indonesia).

Dataset Authors: Nur Fathur Rachman, Moh. Naufal Thariq, Rypaldho Ridotua Hutagaol, Asep Prayogi, and Dr. Cucun Very Angkoso, S.T., M.T.

Unlike CLAHE-enhanced derivatives, all images in this repository are preserved as unenhanced standard sRGB (.jpg, 224 Γ— 224 px) organized directly in raw folder hierarchy (<Species>/<Roast>/*.jpg) rather than pre-packaged Parquet shards, enabling direct inspection, custom preprocessing pipelines (e.g., CIE LAB CLAHE, color-space ablation), or standard ImageFolder loading.


Directory Structure (Raw Files)

coffee-beans-2000-raw/
β”œβ”€β”€ Arabika/
β”‚   β”œβ”€β”€ Dark/      (125 .jpg images)
β”‚   β”œβ”€β”€ Green/     (125 .jpg images)
β”‚   β”œβ”€β”€ Light/     (125 .jpg images)
β”‚   └── Medium/    (125 .jpg images)
β”œβ”€β”€ Ekselsa/
β”‚   β”œβ”€β”€ Dark/      (125 .jpg images)
β”‚   β”œβ”€β”€ Green/     (125 .jpg images)
β”‚   β”œβ”€β”€ Light/     (125 .jpg images)
β”‚   └── Medium/    (125 .jpg images)
β”œβ”€β”€ Liberika/
β”‚   β”œβ”€β”€ Dark/      (125 .jpg images)
β”‚   β”œβ”€β”€ Green/     (125 .jpg images)
β”‚   β”œβ”€β”€ Light/     (125 .jpg images)
β”‚   └── Medium/    (125 .jpg images)
β”œβ”€β”€ Robusta/
β”‚   β”œβ”€β”€ Dark/      (125 .jpg images)
β”‚   β”œβ”€β”€ Green/     (125 .jpg images)
β”‚   β”œβ”€β”€ Light/     (125 .jpg images)
β”‚   └── Medium/    (125 .jpg images)
β”œβ”€β”€ metadata.csv   (2,000 rows: file_name, label, species, roast_level)
└── README.md

Class Distribution

Folder (species) Botanical Taxon Green Light Medium Dark Subtotal
Arabika Coffea arabica 125 125 125 125 500
Ekselsa Coffea liberica var. dewevrei 125 125 125 125 500
Liberika Coffea liberica 125 125 125 125 500
Robusta Coffea canephora 125 125 125 125 500
Total 4 Species Γ— 4 Roast Levels (16 Classes) 500 500 500 500 2,000
  • Format & Resolution: Standard RGB .jpg, 224 Γ— 224 pixels (~24.5 MB total).
  • Preprocessing Status: Raw RGB (no CLAHE contrast enhancement applied).

Usage

1. Download as Raw Folder Structure (Species/Roast/*.jpg)

from huggingface_hub import snapshot_download

local_dir = snapshot_download(
    repo_id="turasa22/coffee-beans-2000-raw",
    repo_type="dataset",
    local_dir="./coffee_beans_2000_raw",
)

Or via Git LFS:

git lfs install
git clone https://huggingface.co/datasets/turasa22/coffee-beans-2000-raw

2. Load via Hugging Face datasets

from datasets import load_dataset

ds = load_dataset("turasa22/coffee-beans-2000-raw")
print(ds["train"][0])
# {'image': <PIL.JpegImagePlugin.JpegImageFile mode=RGB size=224x224>,
#  'label': 'Arabika_Dark', 'species': 'Arabika', 'roast_level': 'Dark'}

Authors & Attribution

  • Dataset Authors:
    • Nur Fathur Rachman (Informatics Engineering, Universitas Trunojoyo Madura)
    • Moh. Naufal Thariq (Informatics Engineering, Universitas Trunojoyo Madura)
    • Rypaldho Ridotua Hutagaol (Informatics Engineering, Universitas Trunojoyo Madura)
    • Asep Prayogi (Informatics Engineering, Universitas Trunojoyo Madura)
    • Dr. Cucun Very Angkoso, S.T., M.T. (Research Supervisor, Informatics Engineering, Universitas Trunojoyo Madura)
  • Affiliation: Informatics Engineering Study Program, Faculty of Engineering, Universitas Trunojoyo Madura (UTM), Bangkalan, Indonesia.
  • Related Repository (Models & 5-Fold Benchmark): turasa22/klasifikasibijikopiconvnext
  • Intellectual Property Record: Directorate General of Intellectual Property (DJKI), Ministry of Law of the Republic of Indonesia β€” Copyright Registration No. 001554369.
  • License: Creative Commons Attribution 4.0 International (CC BY 4.0).
Downloads last month
200