SmartPoultry: Early Detection of Poultry Disease from Smartphone Captured Fecal Image

Abstract

The outbreak of chicken disease has been a major concern around the world, as the poultry industry supplies a significant portion of t he global protein needs. Such an outbreak can cause enormous financial loss to the poultry farmers and induce food insecurity. The COVID-19 lessons have taught us that chicken disease outbreak can be a threat to human lives as well if not detected in time. Currently, Poultry farmers rely on their experience to detect diseases and to seek professional's help, which occasionally fails, resulting in widespread chicken death. Thus, early detection of chicken disease is of great importance for sustainable poultry farming, reducing poultry losses and preventing the spread of zoonotic diseases to humans. Several methods proposed previously for this purpose have failed to achieve sufficient a ccuracy and practical usability. In this paper, we present an AI-assisted automated system for detecting chicken diseases at an early stage from smart-phone captured fecal images. The proposed method utilized an ensemble network of four fine-tuned convolutional neural networks that were selected through an exhaustive literature search. The proposed method outperformed existing methods, achieving 99.99% accuracy and we demonstrated its practical usability in terms of time, robustness, user friendliness and cost.