Abstract

Disease detection in poultry farming is crucial for ensuring the health and productivity of chickens. This paper presents a novel deep learning-based approach for automated disease classification using images of chicken fecal matter. The objective is to develop a model capable of accurately identifying diseases early, facilitating timely intervention and effective management in poultry farms. The methodology encompasses several key components. Firstly, a diverse and balanced dataset of chicken fecal samples is collected, comprising images representing various diseases. Data preprocessing techniques such as image normalization, resizing, and augmentation are applied to enhance the model's robustness and generalization ability. The neural network architecture utilizes transfer learning with MobileNetV3Small, a lightweight yet powerful convolutional neural network (CNN) architecture known for its efficiency in image classification tasks. Results indicate that the developed model achieves competitive test accuracy, demonstrating its effectiveness in classifying diseases in chicken fecal samples. Precision, recall, and F1-score metrics further illustrate the model's ability to correctly identify and differentiate between different disease classes. The use of transfer learning with MobileNetV3Small contributes to the model's success by leveraging pre-trained features and optimizing computational efficiency.