Good and Bad Classification of Parotta

Description:

The project titled "Good and Bad Classification of Parotta Using Realme P1 5G Mobile Camera" focuses on developing an image-based machine learning system capable of automatically classifying parotta samples into two categories: good quality parotta and bad quality parotta. The objective of this project is to assist in food quality assessment by using computer vision techniques to identify freshness, appearance, texture, and visible defects in parotta samples. This system can help food industries, restaurants, bakeries, and consumers maintain quality standards and reduce food wastage.

The dataset used in this project consists of more than 1000 images, with over 500 images of good-quality parotta and 500 images of bad-quality parotta. The images were captured using the Realme P1 5G smartphone, which is equipped with a 50 MP AI rear camera capable of capturing high-resolution images with excellent detail. The high-quality camera allows clear visualization of important features such as color, texture, surface condition, and shape of the parotta.

Dataset Composition:

Good Samples (Fresh and High-Quality Parotta):
More than 500 images represent good-quality parotta. These samples exhibit desirable characteristics such as an even golden-brown color, soft texture, proper layering, uniform shape, and absence of burns, cracks, mold, or contamination. These images form the positive class and represent freshly prepared parotta suitable for consumption.

Bad Samples (Poor-Quality or Spoiled Parotta):
The dataset also contains more than 500 images of bad-quality parotta. These samples may show signs of spoilage, staleness, excessive dryness, hard texture, burnt surfaces, discoloration, fungal growth, contamination, or improper preparation. These images form the negative class and help the model learn to identify defective or unsafe parotta samples.
Data Collection Setup:
Images were captured using the Realme P1 5G mobile camera under controlled conditions. A black background was used to create strong contrast between the parotta and the surroundings, enabling easier feature extraction. The photographs were taken under natural daylight conditions supplemented with 360-degree LED lighting, ensuring uniform illumination and reducing shadows or lighting variations. This setup improved image consistency and enhanced the visibility of surface features.
Image Characteristics:
The dataset includes parotta samples with variations in size, thickness, shape, texture, and preparation methods. Such diversity is important for training a robust machine learning model capable of accurately classifying different types of parotta under real-world conditions. The images capture a wide range of quality levels, making the dataset suitable for practical food quality inspection applications.
Data Annotation:
Each image was manually labeled as either "Good" or "Bad" based on visual quality assessment and expert observation.