This dataset contains a large-scale benchmark of 240 broilers with a total of 9,214 high-resolution images (1280×720, PNG format). It is designed for broiler pose estimation and lameness detection tasks, and includes manually annotated anatomical key points and a defined pose skeleton. All annotations are stored in Comma-Separated Values (CSV) format.

The dataset was created to support research in:

• Animal welfare monitoring.

• Computer vision–based lameness assessment.

• Pose estimation and keypoint detection.

📍 Keypoint Definitions & Pose Skeleton

Each broiler image is annotated with 7 anatomical key points:

🗂️ Data Structure

📁 Lameness/

│

└── 📁 Broiler_ID/

├── 📄 img50.png

├── 📄 img52.png

└── ...

├── 📄 CollectedData.csv

└── ...

└── 📁 Broiler_ID/

├── 📄 img15.png

├── 📄 img17.png

└── ...

├── 📄 CollectedData.csv

📝 Annotation Format

Each CollectedData.csv file has the following structure:

First row: bodyparts -- lists all annotated key points (each with x and y columns).

Second row: coords -- specifies the coordinate axes (x, y).

Subsequent rows: One row per image, starting with the file name followed by coordinates for each key point.

Note: Missing values indicate that the key point is not visible in the image.

📑 Data Summary

📥 Dataset Download

You can directly download the dataset from here:

📚 Citation

If you use this dataset, please cite:

Nasiri, A., Yoder, J., Zhao, Y., Hawkins, S., Prado, M. and Gan, H., 2022. Pose estimation-based lameness recognition in broiler using CNN-LSTM network. Computers and Electronics in Agriculture, 197, p.106931.

DOI: https://doi.org/10.1016/j.compag.2022.106931

🤝 Contributing

Contributions are welcome!