Bad and Good classifing Momo

The project titled “Good and Bad Classification of Momo Using Infinix GT 20 Pro Mobile Camera” focuses on developing a machine learning and computer vision system to automatically classify momo samples into two categories: good (fresh and high-quality) and bad (stale, spoiled, or poor-quality). The objective of this project is to build an accurate image classification model capable of distinguishing fresh momos from deteriorated or contaminated ones using high-quality image data.
The dataset consists of more than 1000 images, including over 500 good-quality momo images and 500 bad-quality momo images. All images were captured using the Infinix GT 20 Pro smartphone, featuring a 108 MP Samsung ISOCELL HM6 primary camera with a 1/1.67-inch sensor, f/1.75 aperture, Optical Image Stabilization (OIS), and Electronic Image Stabilization (EIS). The device also includes 2 MP macro and 2 MP depth sensors, enabling detailed image capture. The high-resolution camera preserves important visual features such as texture, color, shape, folds, and surface defects, which are essential for accurate classification.
Dataset Composition:
Good Samples (Fresh Momo):
The dataset contains more than 500 images of fresh and high-quality momos. These samples exhibit desirable characteristics such as a smooth and uniform surface, proper shape, fresh dough appearance, appropriate color, and no visible signs of spoilage or damage. These images represent the positive class and define the standard of acceptable product quality.
Bad Samples (Poor-Quality Momo):
The dataset also includes over 500 images of poor-quality momos. These samples may show signs of spoilage, fungal growth, discoloration, broken structure, excessive dryness, contamination, or improper cooking. These images form the negative class and help the model learn to recognize defective or unsafe food products.
Data Collection Setup:
Images were captured under controlled conditions using the Infinix GT 20 Pro mobile camera. A white background was used to create a clean and uniform contrast between the momo samples and the surroundings, making feature extraction more effective. Images were captured under natural daylight with additional 360-degree LED lighting to ensure consistent illumination while minimizing shadows and reflections. This setup resulted in a reliable and high-quality dataset.
Image Characteristics:
The dataset includes momo samples with variations in size, shape, folding style, texture, color, and quality conditions. This diversity improves the robustness of the machine learning model and enables it to generalize effectively to different types of momos encountered in real-world situations.
Data Annotation:
Each image was manually labeled as either “Good” or “Bad” based on visual quality assessment by expert observation. These annotations serve as ground truth labels for supervised learning, ensuring accurate model training and evaluation.

本项目题为《基于Infinix GT 20 Pro移动摄像头的馍馍（momo）优劣分类》，旨在开发一套机器学习与计算机视觉系统，实现对馍馍样本的自动分类，将其划分为两类：优质（新鲜且品质优良）与劣质（过期、变质或品质低劣）。本项目的目标是构建精准的图像分类模型，借助高质量图像数据，区分新鲜馍馍与变质或受污染的馍馍。

本数据集包含超过1000张图像，其中优质馍馍图像超500张，劣质馍馍图像超500张。所有图像均采用Infinix GT 20 Pro智能手机拍摄，该设备搭载108 MP三星ISOCELL HM6主摄，配备1/1.67英寸传感器、f/1.75光圈、光学图像防抖（Optical Image Stabilization，OIS）及电子图像防抖（Electronic Image Stabilization，EIS）；此外还内置2 MP微距传感器与2 MP深度传感器，可实现精细图像采集。这款高分辨率摄像头能够保留纹理、色泽、形状、褶皱及表面缺陷等关键视觉特征，这些特征对精准分类至关重要。

数据集构成：
优质样本（新鲜馍馍）：
本数据集包含超500张新鲜优质馍馍的图像。此类样本具备理想品质特征：表面光滑均匀、形状规整、面皮质感新鲜、色泽适宜，无可见变质或损伤痕迹。这些图像作为正样本，定义了合格产品的品质标准。

劣质样本（品质低劣馍馍）：
本数据集还包含超500张品质低劣馍馍的图像。此类样本可能出现变质、霉菌滋生、变色、结构破损、过度干燥、受污染或烹饪不当等问题。这些图像作为负样本，帮助模型学习识别缺陷或不安全的食品。

数据采集设置：
所有图像均采用Infinix GT 20 Pro移动摄像头在可控环境下拍摄。采集时采用白色背景，使馍馍样本与背景形成清晰均匀的对比度，提升特征提取效率；同时使用自然光搭配360度LED补光灯，确保光照均匀，最大限度减少阴影与反光。该采集方案可生成可靠且高质量的数据集。

图像特征：
本数据集涵盖尺寸、形状、包制手法、纹理、色泽及品质状态存在差异的馍馍样本。这种多样性可提升机器学习模型的鲁棒性，使其能够有效泛化至实际场景中遇到的各类馍馍。

数据标注：
所有图像均由专家通过视觉品质评估，手动标注为“优质”或“劣质”。这些标注作为监督学习的真值标签，为模型的精准训练与评估提供保障。