Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Lightweight Model for Poultry Disease Detection from Fecal Images Using Multi-Color Space Feature Optimization and Machine Learning

A. K. M. Shoriful Islam, Md. Rakib Hassan, Macbah Uddin, Md. Shahidur Rahman

TL;DR

The paper addresses the need for real-time, low-resource poultry disease detection from fecal images. It proposes a lightweight pipeline that leverages multi-color-space features (RGB, HSV, LAB) and uses PCA and XGBoost for feature selection, followed by an artificial neural network classifier. A carefully engineered Global Feature Set (LAB-CM, HSV-LBP, LAB-LBP) achieves up to 95.60–95.85% accuracy on a diverse dataset, with CPU-only execution and substantially lower resource requirements than CNN-based approaches. The work demonstrates a cost-effective, interpretable alternative suitable for deployment in small-scale, low-resource poultry farming contexts, with clear avenues for mobile/edge deployment and further expansion to more diseases.

Abstract

Poultry farming is a vital component of the global food supply chain, yet it remains highly vulnerable to infectious diseases such as coccidiosis, salmonellosis, and Newcastle disease. This study proposes a lightweight machine learning-based approach to detect these diseases by analyzing poultry fecal images. We utilize multi-color space feature extraction (RGB, HSV, LAB) and explore a wide range of color, texture, and shape-based descriptors, including color histograms, local binary patterns (LBP), wavelet transforms, and edge detectors. Through a systematic ablation study and dimensionality reduction using PCA and XGBoost feature selection, we identify a compact global feature set that balances accuracy and computational efficiency. An artificial neural network (ANN) classifier trained on these features achieved 95.85% accuracy while requiring no GPU and only 638 seconds of execution time in Google Colab. Compared to deep learning models such as Xception and MobileNetV3, our proposed model offers comparable accuracy with drastically lower resource usage. This work demonstrates a cost-effective, interpretable, and scalable alternative to deep learning for real-time poultry disease detection in low-resource agricultural settings.

Lightweight Model for Poultry Disease Detection from Fecal Images Using Multi-Color Space Feature Optimization and Machine Learning

TL;DR

The paper addresses the need for real-time, low-resource poultry disease detection from fecal images. It proposes a lightweight pipeline that leverages multi-color-space features (RGB, HSV, LAB) and uses PCA and XGBoost for feature selection, followed by an artificial neural network classifier. A carefully engineered Global Feature Set (LAB-CM, HSV-LBP, LAB-LBP) achieves up to 95.60–95.85% accuracy on a diverse dataset, with CPU-only execution and substantially lower resource requirements than CNN-based approaches. The work demonstrates a cost-effective, interpretable alternative suitable for deployment in small-scale, low-resource poultry farming contexts, with clear avenues for mobile/edge deployment and further expansion to more diseases.

Abstract

Poultry farming is a vital component of the global food supply chain, yet it remains highly vulnerable to infectious diseases such as coccidiosis, salmonellosis, and Newcastle disease. This study proposes a lightweight machine learning-based approach to detect these diseases by analyzing poultry fecal images. We utilize multi-color space feature extraction (RGB, HSV, LAB) and explore a wide range of color, texture, and shape-based descriptors, including color histograms, local binary patterns (LBP), wavelet transforms, and edge detectors. Through a systematic ablation study and dimensionality reduction using PCA and XGBoost feature selection, we identify a compact global feature set that balances accuracy and computational efficiency. An artificial neural network (ANN) classifier trained on these features achieved 95.85% accuracy while requiring no GPU and only 638 seconds of execution time in Google Colab. Compared to deep learning models such as Xception and MobileNetV3, our proposed model offers comparable accuracy with drastically lower resource usage. This work demonstrates a cost-effective, interpretable, and scalable alternative to deep learning for real-time poultry disease detection in low-resource agricultural settings.

Paper Structure

This paper contains 21 sections, 6 figures, 13 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Working Procedure
  4. Dataset Description
  5. Image Mode Conversion
  6. Feature Extraction
  7. Classification Algorithms
  8. Hardware description
  9. Experimental Results
  10. Feature Selection From Color Spaces
  11. Combination of Features From Different Color Spaces
  12. Color Feature Optimization and Exclusion Justification
  13. Texture Feature Optimization and Exclusion Justification
  14. Feature Optimization and Exclusion Justification in Shape Analysis
  15. Color + Texture Feature Set Optimization
  16. ...and 6 more sections

Figures (6)

  • Figure 1: Proposed Model
  • Figure 2: Color space conversion from RGB image
  • Figure 3: Accuracy comparison on color channels
  • Figure 4: Accuracy comparison color, shape and texture analysis
  • Figure 5: Confusion Matrix
  • ...and 1 more figures