MANGO PICKLE

Project Title: Good and Bad Classification of Mango Pickle (Mangifera indica) Using Redmi 10 Prime Mobile Camera

Description:

The project, "Good and Bad Classification of Mango Pickle (Mangifera indica)," aims to develop a machine learning model capable of distinguishing between good and bad samples of mango pickle. A dataset of more than 500 images, evenly divided between good and bad mango pickle samples, has been created. The images were captured using a Redmi 10 Prime mobile camera, providing high-resolution visual data. A white background was used to highlight the pickle, and images were taken in natural daylight to ensure optimal lighting conditions for accurate analysis.

Dataset Composition:

Good Samples (High-Quality Mango Pickle): The dataset includes more than 250 images of high-quality mango pickle. These samples exhibit the desired characteristics of a good pickle, such as vibrant color, proper texture, correct oil and spice distribution, and absence of spoilage indicators like mold or off-putting discoloration. These images represent the positive class in the model, providing examples of what constitutes a well-made, high-quality mango pickle.

Bad Samples (Low-Quality Mango Pickle): Another 250 images represent low-quality or spoiled mango pickle. These samples may show signs of spoilage such as mold, fermentation bubbles, discoloration, or improper texture (either too dry or too watery). These images make up the negative class, helping the model learn to identify pickle samples that do not meet quality standards.

Data Collection Setup:

The images were captured using the Redmi 10 Prime mobile camera, chosen for its ability to capture high-quality, detailed images. The use of a white background helped to create a clear, distraction-free environment, ensuring that the model focuses on the key features of the pickle. Natural daylight was used for illumination to bring out the true colors and textures, providing consistency and reducing shadows or reflections that could distort the image quality.

Image Characteristics:

The images in the dataset vary in terms of the appearance of the mango pickle, influenced by different ingredients, preparation methods, and levels of spoilage or aging. This variability allows the model to generalize well and perform effectively in different scenarios, whether identifying a batch of well-preserved pickles or detecting signs of spoilage.

Data Annotation:

Each image has been manually labeled as either "good" or "bad" based on expert observation of the pickle’s quality. These labels serve as the ground truth, allowing the model to learn the differences between high-quality and low-quality pickle samples.

Data Preprocessing:

The dataset is processed through several steps to ensure that the machine learning model receives clean, standardized inputs:

Resizing: All images are resized to a uniform dimension to ensure consistency across the dataset.

项目名称：基于Redmi 10 Prime手机摄像头的芒果酱菜（Mangifera indica）优劣分类

项目简介：
本项目旨在开发可区分优质与劣质芒果酱菜样本的机器学习模型。本次构建的数据集包含超500张图像，优质与劣质样本数量均分。所有图像均通过Redmi 10 Prime手机摄像头拍摄，获取高分辨率视觉数据；拍摄时采用纯白背景以突出酱菜主体，并使用自然日光以确保最佳光照条件，保障后续分析的准确性。

数据集构成：
优质样本（高品质芒果酱菜）：数据集包含超250张高品质芒果酱菜图像。此类样本具备优质酱菜的理想特征，如色泽鲜亮、质地规整、油分与香料分布均匀，且无霉变、异常变色等变质迹象。这些图像作为模型的正样本，用于提供合格优质芒果酱菜的判定依据。
劣质样本（低品质芒果酱菜）：另有250张图像对应低品质或已变质的芒果酱菜。此类样本可观察到变质迹象，例如霉变、发酵气泡、色泽异常，或质地不当（过干或过稀）。这些图像作为负样本，帮助模型学习识别不符合质量标准的酱菜样本。

数据采集设置：
图像采集采用Redmi 10 Prime手机摄像头，该设备可拍摄高质量、细节丰富的图像。纯白背景的使用能够打造清晰无干扰的拍摄环境，确保模型聚焦于酱菜的核心特征。拍摄时采用自然日光照明，可还原食材真实色泽与质地，保证光照一致性，减少可能影响图像质量的阴影与反光。

图像特征：
本数据集内的图像因食材配比、制作工艺、变质程度与陈化时长的差异而呈现多样化外观。这种多样性有助于模型提升泛化能力，使其在各类场景下均可有效运作，无论是识别批量保质完好的酱菜，还是检测变质迹象。

数据标注：
所有图像均经人工标注为“优质”或“劣质”，标注依据为专业人员对酱菜品质的观测结果。这些标注作为模型训练的真实标签（ground truth），使模型能够学习优质与劣质酱菜样本间的差异。

数据预处理：
为确保机器学习模型获得干净且标准化的输入数据，数据集将经过以下预处理步骤：
1. 尺寸统一：将所有图像调整至统一尺寸，以保证数据集内样本的一致性。