Pigs: A stepwise RGB-D novel pig carcass cutting dataset.

Area of Science:

• Robotics and Computer Vision
• Agricultural Technology
• Food Processing Automation

Background:

• Accurate 3D data is crucial for automated systems in food processing.
• Existing datasets may not fully capture the complexities of carcass manipulation.
• Developing robust datasets is essential for advancing robotic applications in agriculture.

Purpose of the Study:

• To present a comprehensive dataset for pig carcass cutting applications.
• To provide essential data for training and validating computer vision algorithms in meat processing.
• To facilitate the development of advanced robotic systems for the meat industry.

Main Methods:

• Data acquisition using a custom frame with 6 Intel® RealSense™ Depth Camera D415 units.
• Recording data via a robotic arm equipped with a single camera, replicating frame positions.
• Data storage in ROS bag files, including RGB-D streams and camera intrinsic parameters.
• Provision of transformation matrices in JSON format for precise spatial calibration.

Main Results:

• A detailed RGB-D dataset of pig carcasses suitable for cutting tasks.
• Accurate camera intrinsic and extrinsic parameters for each sensor.
• Transformation matrices enabling accurate 3D reconstruction and robotic path planning.
• Two distinct data capture methodologies (static frame and robotic arm) for diverse application needs.

Conclusions:

• The presented dataset offers a valuable resource for research in agricultural robotics and food processing.
• The inclusion of detailed spatial information supports the development of precise automated cutting systems.
• This dataset can accelerate innovation in automated meat handling and processing technologies.