Agri-Human Dataset Tools

This main branch consolidates the code that previously lived across multiple branches into a single, organized layout while leaving the original branches untouched.

Layout

.
|-- LICENSE
|-- README.md
|-- shared/
|   |-- README.md
|   |-- build_manifest_and_splits.py
|   `-- sync_and_match.py
|-- kitti/
|   |-- README.md
|   |-- kitti_export.yaml
|   |-- kitti_export_common.py
|   |-- kitti_export_ctl.py
|   |-- kitti_export_custom.py
|   |-- kitti_export_depth.py
|   |-- kitti_export_object.py
|   `-- kitti_export_raw.py
|-- yolo/
|   |-- README.md
|   |-- coco_export_session.py
|   `-- yolo_export_session.py
|-- converters/
|   |-- fieldsafe_rgb_to_yolo.py
|   `-- yolo_to_coco.py
|-- filters/
|   `-- filter_to_person.py
|-- preprocessing/
|   `-- undistort_dataset_images.py
`-- ros2bag/
    |-- README.md
    `-- check_and_make_rosbag2.py

Branch Origins

• shared/ comes from v1.0 and contains the common dataset synchronization and manifest tooling.
• kitti/ comes from KITTI-converter.
• yolo/ comes from YOLO-converter.
• converters/ contains reusable format-transform utilities.
• filters/ contains dataset filtering utilities.
• preprocessing/ contains dataset preparation scripts such as image undistortion.
• ros2bag/ comes from ROS2bag-converter.

The original branches remain available as-is:

• KITTI-converter
• origin/KITTI-converter
• origin/YOLO-converter
• origin/ROS2bag-converter
• origin/v1.0

How To Use This Branch

You can either run scripts from the repository root with explicit paths:

python .\shared\sync_and_match.py <DATASET_ROOT>
python .\shared\build_manifest_and_splits.py --root <DATASET_ROOT>
python .\kitti\kitti_export_object.py --root <DATASET_ROOT> --out <OUT_DIR>
python .\yolo\yolo_export_session.py --root <DATASET_ROOT> --out <OUT_DIR>
python .\converters\yolo_to_coco.py --images_dir <IMAGES_DIR> --labels_dir <LABELS_DIR> --yaml_path <DATA_YAML> --output_path <OUT_JSON>
python .\filters\filter_to_person.py --format coco --ann_file <ANN_JSON> --img_dir <IMG_DIR> --output <OUT_JSON> --human_labels person
python .\preprocessing\undistort_dataset_images.py --root <DATASET_ROOT> --intrinsics <INTRINSICS_JSON>
python .\ros2bag\check_and_make_rosbag2.py --bag-dir <DATASET_BAG_DIR>

Or cd into a toolkit folder and follow the local README.md there.

For AGHRI ZED RGB/Livox ROS 2 playback, ros2bag/check_and_make_rosbag2.py is the canonical converter. Its calibration-aware mode reads calibration/intrinsics.json and calibration/extrinsics.json and writes standard ZED/fisheye CameraInfo topics, /tf, and /tf_static messages into the bag without changing the original image or LiDAR timestamps.

Notes

• shared/build_manifest_and_splits.py and shared/sync_and_match.py are identical across v1.0, KITTI-converter, and YOLO-converter, so they are stored once.
• Each toolkit keeps its own branch-specific README.md to preserve the original usage guidance close to the code.