Chicken Health and Behavior Multimodal Dataset (chicken-health-behavior-multimodal)

This dataset provides a comprehensive collection of visual and audio data from chicken farms, specifically designed for early detection of chicken health issues and anomalous behaviors. It aims to support the development of intelligent monitoring systems that can mitigate significant economic losses caused by poultry disease, with a particular focus on future applications.

Why this Dataset is Crucial

The poultry sector in Indonesia is a vital economic pillar, with substantial domestic investment (IDR 1.8 trillion in 2021). However, this industry faces severe threats from disease outbreaks, which lead to substantial financial losses and jeopardize business sustainability.

Disease like Newcastle Disease (ND), a highly contagious viral infection, can cause up to 100% morbidity and mortality, with economic losses reaching IDR 474,500,000.00 in a single farm over six month. Chronic Respiratory Disease (CRD), caused by Mycoplasma gallisepticum, also inflicts significant economic damage, potentially reducing egg production by over 50% and incurring monthly losses of IDR 342,114.00 per 650 chickens, representing a 32.52% profit reduction. Complications from CRD can escalate mortality to 30%.

Common symptoms of sick chickens are varied and often hard for farmers to detect early through manual inspection. These include reduced appetite, respiratory distress, sneezing, weight loss, dull feathers, decreased egg production, various types of diarrhea, pale or bluish faces, swollen or pale combs, lameness, eye/nasal discharge, swollen or twisted heads, watery eyes, drooping wings, soiled vents, resting with beaks on the floor, crouching, soft bodies, open navels, warts, stunted growth, falling sideways, head/neck tremors, ruffled feathers, pale skin/beaks, mucoid or wormy diarrhea, seizures, paralysis, lethargy, huddling, and significantly below-average weight. Manual monitoring methods are prone to fatigue, decreased concentration, and burnout, leading to misjudgments. Lack of access to modern technology and high operational costs also pose significant barriers for small-scale farmers.

The combined high mortality rates (up to 100% for ND, 30% for complicated CRD) and massive financial losses (hundreds of millions to billions of rupiah annually per farm) pose an existential threat to the profitability and sustainability of Indonesian poultry farming. This situation demands technological intervention to ensure industry stability, food security, and the livelihoods of many individuals. Beyond direct losses, the inability to effectively monitor and prevent diseases also incurs substantial opportunity costs in unrealized production potential, hindering farm scalability, improved practices, and animal welfare.

This dataset directly addresses this critical need by providing the data required to build innovative systems for early detection and accurate insights, aiming to significantly reduce losses and boost productivity.

Dataset Structure and Content

The chicken-health-behavior-multimodal dataset is designed to serve different research and application needs, encompassing both computer vision and vocalization analysis.

Data Source :

• Video/Image Data: The visual data (images/videos) included in the vision-object-detection configuration has been curated from publicly available videos, specifically from the Kipster Farm YouTube channel. While Kipster Farm is located in the US, this data is utilized for its high-quality depiction of chicken behavior and general farm environments, which are relevant for developing foundational detection and tracking models applicable to various poultry farming contexts. We acknowledge that specific environmental or breed differences may exist, and future expansions aim to include data directly from Indonesian farms.

• Audio Data: The audio dataset is sourced from the "Poultry Vocalization Signal Dataset for Early Disease Detection". The aim of this dataset is to deliver open, accessible, and quality machine learning data for poultry farm management.

Data Structure :

audio : This configuration contains audio clips of chicken vocalizations with corresponding labels for health or behavioral states. images : Still images of chickens in various farm environments. videos : Video clips with accompanying synchronized audio streams.

Acknowledgements

• We extend our gratitude to the poultry farmers and researchers whose insights and data contribute to this project.
• Audio Data : Aworinde, Halleluyah; Adebayo, Segun; Akinwunmi, Akinwale; Alabi, Olufemi; Ayandiji, Adebamiji; Oke, Olaide; Oyebamiji, Abel; Adeyemo, Adetoye; Sakpere, Aderonke; Echetama, Kizito (2023), “Poultry Vocalization Signal Dataset for Early Disease Detection”, Mendeley Data, V1.
• Video Data : youtube.com/@kipsterfarm and kipster.farm

Contribution and Contact

We welcome contributions and feedback to improve this dataset. If you have any questions or suggestions, please open an issue on the dataset repository.