README.md

metadata

license: mit
task_categories:
  - object-detection
  - robotics
language:
  - en
tags:
  - Robotics,
  - AI,
pretty_name: FAFO
size_categories:
  - 10K<n<100K

FAFO Dataset

The FAFO dataset is designed for universal robotics software development. It includes:

• Sensor data: LiDAR scans, GPS coordinates, and IMU readings.
• Image data: Infrared and camera images for object detection and navigation.
• 3D data: Point cloud files for SLAM and mapping.
• Task data: Pre-labeled tasks for robotic arm operations.

Dataset Overview

Sensor Data

• LiDAR Data: Point cloud scans with timestamps, ranges, intensities, and angles
• GPS Data: Precise location data including latitude, longitude, and altitude
• IMU Data: Acceleration, angular velocity, and orientation readings

Image Data

• RGB camera feeds
• Infrared images
• Object detection datasets with bounding box annotations

3D Data

• Point cloud maps for SLAM
• 3D environment scans
• Occupancy grid maps

Task Data

• Pick-and-place task definitions
• Navigation paths
• Robot arm trajectories
• Task annotations and metadata

Dataset Structure

• sensor_data/: Contains JSON files for LiDAR, GPS, and IMU readings.
• image_data/: JPEG images for object detection and segmentation.
• 3d_data/: PCD files for 3D point clouds.
• task_data/: JSON files for robotic tasks.

Usage

This dataset is designed for AI model training, sensor calibration, and robotic task automation.

Loading the Dataset

from datasets import load_dataset

# Load the complete dataset
dataset = load_dataset("GotThatData/fafo")

# Load specific splits
train_dataset = load_dataset("GotThatData/fafo", split="train")
val_dataset = load_dataset("GotThatData/fafo", split="validation")
test_dataset = load_dataset("GotThatData/fafo", split="test")

Example Usage

# Access sensor data
lidar_scan = dataset['train'][0]['sensor_data']['lidar']
gps_reading = dataset['train'][0]['sensor_data']['gps']
imu_data = dataset['train'][0]['sensor_data']['imu']

# Access image data
image_path = dataset['train'][0]['image_data']

# Access 3D data
point_cloud = dataset['train'][0]['3d_data']

# Access task data
task = dataset['train'][0]['task_data']

Data Format

Sensor Data

{
    "lidar": {
        "timestamp": 1640995200.0,
        "ranges": [1.2, 2.3, 3.4],
        "intensities": [0.5, 0.6, 0.7],
        "angles": [0.0, 0.1, 0.2]
    },
    "gps": {
        "timestamp": 1640995200.0,
        "latitude": 37.7749,
        "longitude": -122.4194,
        "altitude": 0.0
    },
    "imu": {
        "timestamp": 1640995200.0,
        "acceleration": [0.0, 0.0, 9.81],
        "angular_velocity": [0.0, 0.0, 0.0],
        "orientation": [0.0, 0.0, 0.0, 1.0]
    }
}

Task Data

{
    "task_type": "pick_and_place",
    "parameters": {
        "position": [0.5, 0.3, 0.2],
        "orientation": [0.0, 0.0, 0.0, 1.0],
        "gripper_state": "open"
    },
    "annotations": {
        "object_class": "cube",
        "bounding_box": [0.1, 0.1, 0.2, 0.2],
        "confidence": 0.95
    }
}

Dataset Statistics

• Total samples: [Number of samples]
• Train/Val/Test split: 60%/20%/20%
• Data types:
  • Sensor readings: [Number of readings]
  • Images: [Number of images]
  • 3D scans: [Number of scans]
  • Task definitions: [Number of tasks]

License

MIT License

Citation

@inproceedings{fafo2024,
    title={FAFO Dataset},
    author={GotThatData},
    year={2024}
}