Fruit Quality – Good vs Bad for Apple, Banana, Lime

This dataset contains colour images of top fruit types with two quality labels: Good and Bad. It includes six classes: Apple_Good, Apple_Bad, Banana_Good, Banana_Bad, Lime_Good, Lime_Bad. The approximate sample counts per class are:

• Apple_Bad: 1141

• Apple_Good: 1134

• Banana_Bad: 1087

• Banana_Good: 1113

• Lime_Bad: 1085

• Lime_Good: 1094

The images were captured under varying lighting conditions, backgrounds and viewpoints, using high-resolution mobile phone cameras, both indoor and outdoor. The dataset is derived from the FruitNet dataset (Meshram et al., 2022) which originally included six fruits (apple, banana, guava, lime, orange, pomegranate) and three quality labels (Good/Bad/Mixed) with ~19 500 images. 

In this version we have selected the six classes (three fruits × two quality levels) to provide a balanced corpus of ~ 6,700 images. Each image is stored at 256×256 resolution (or rescaled to this size) and labelled with the class name.

Intended Use:
This dataset is suitable for research in computer vision, particularly fruit quality classification, defect detection, visual sorting in agriculture, and related machine learning tasks.

Source / Reference:
Meshram V. A., Patil K., Ramteke S. D. “MNet: A Framework to Reduce Fruit Image Misclassification” (2021) IIETA. DOI: 10.18280/isi.260203. Dataset details: top Indian fruits, Good/Bad labels, ~12,000 images. (IIETA)

Dataset structuring:

• Directory per class (e.g. Apple_Bad/)

• JPEG or PNG images

• Recommended preprocessing: resizing to 256×256, normalising pixel values, optional data augmentation