Lojano Arabica coffee

This dataset contains a collection of Loja Arabica green coffee bean images captured for research in automated quality assessment and binary classification of coffee beans (good vs. defective). The dataset is structured to support both raw image analysis and model training using preprocessed, uniformly sized samples.

Contents

The dataset is organized into two main parts:

1. Raw Images

Original high-resolution photographs of green coffee beans.

Files available in RAW and JPG formats.

Images contain multiple beans per frame under controlled lighting and a uniform background.

2. Cropped and Classified Images (256 × 256 px)

Individually extracted bean images resized to 256×256 pixels.

Stored in separate folders according to their class label:

good/ — beans without visible defects

defective/ — beans exhibiting defects such as discoloration, deformities, or surface damage

Purpose

This dataset supports research in:

Binary image classification (good vs. defective beans)

Computer vision–based quality control

Deep learning model development (CNNs, ViTs, etc.)

Preprocessing and segmentation studies on agricultural products

Acquisition

Raw images were captured using a fixed-position camera and controlled illumination to minimize shadowing and reflections. Beans were collected from Coffea arabica lots produced in Loja, Ecuador, a region known for its high-quality specialty coffee.

Format

RAW images: original sensor data

JPG images: high-resolution photographs

Cropped images (256×256): JPG or PNG depending on preprocessing