Shelf life determination of Hypophthalmicthys molitrix ( Silver Carp )

This project, titled “Shelf Life Determination of Silver Carp (Hypophthalmichthys molitrix) Using Image Classification,” is designed to develop an image classification system that distinguishes between healthy (good) and unhealthy (bad) Silver Carp fish. The dataset consists of 2000 images, evenly distributed between good and bad samples. All images were captured using a Realme 5i mobile camera, providing high-resolution visual data suitable for machine learning applications.
The fish were photographed against a black background using natural sunlight and LED lighting to ensure uniform illumination, minimize shadows, and improve image clarity.
Dataset Composition
Good Samples (Healthy Fish)
The dataset contains approximately 1000 images of healthy Silver Carp fish. These images show fish with the following characteristics:
Bright and shiny scales
Clear, transparent eyes
Firm body texture
Natural coloration
No visible physical damage, deformity, or spoilage
These images represent the fresh and acceptable condition of fish and are used as positive samples for training the classification model.
Bad Samples (Unhealthy or Spoiled Fish)
The dataset also includes approximately 1000 images of unhealthy or spoiled Silver Carp fish. These fish may show:
Dull or faded scales
Cloudy or sunken eyes
Discoloration
Physical damage or deformities
Signs of spoilage and reduced freshness
These images represent the deteriorated condition of fish and serve as negative samples for training the model.
Data Collection Setup
Images were captured using the Realme 5i smartphone camera, which provides high-quality imaging suitable for detailed visual analysis.
Camera Specifications (Realme 5i)
Main Camera: 12 MP primary sensor with f/1.8 aperture, 5P lens, and Phase Detection Autofocus (PDAF)
Ultra-Wide Lens: 8 MP sensor with 119° field of view and f/2.25 aperture
Macro Lens: 2 MP sensor with 4 cm close-up shooting distance
Portrait Lens: 2 MP depth sensor for background blur
Zoom Capability: Up to 10× digital zoom
Battery: A large 5,000mAh battery .
Charging: Includes a standard 10W (5V/2A) charger.
These camera features allow clear capture of fish texture, color, and structural details essential for classification.
Lighting and Background Conditions
To ensure consistency and high image quality, the following setup was used:
Black background to improve contrast and isolate fish features
Natural sunlight to capture true color and natural appearance
360-degree LED lighting to provide uniform illumination and eliminate shadows
This setup improves feature extraction and enhances model performance.
Image Characteristics
The dataset includes variations in:
Fish size
Body shape
Color intensity
Freshness level
Physical condition
This diversity helps the machine learning model learn robust features and improves its ability to generalize to new data.
Data Annotation
Each image is carefully labeled into one of two categories:
Good (Fresh / Healthy)
Bad (Spoiled / Unhealthy)