ts-polyline-bags

Raw ROS1 bags from the TerraSentia (wheeled robot) used to build the processed dataset adipotnis/ts-polyline.

Public but gated — manual access approval. Click "Request access" on this page; it usually clears within a day.

Contents

22 bags, ~154 GB total (1.5 – 10 GB per file). Bag stems match the traj_id values in adipotnis/ts-polyline. Topics of interest:

Topic	Type	Notes
/terrasentia/zed2/zed_node/left/image_rect_color/compressed	sensor_msgs/CompressedImage	rectified left ZED frame
/terrasentia/zed2/zed_node/left/camera_info	sensor_msgs/CameraInfo	intrinsics (rectified — no distortion)
/terrasentia/ekf	nav_msgs/Odometry	fused EKF pose @ ~100 Hz
/terrasentia/zed2/zed_node/imu/data	sensor_msgs/Imu	ZED IMU @ ~200 Hz
/terrasentia/imu	sensor_msgs/Imu	base IMU
/terrasentia/zed2/zed_node/odom	nav_msgs/Odometry	ZED visual odometry

Use msg.header.stamp for ekf and IMU streams — the rosbag receipt timestamps are batched in microseconds and produce nonsense velocities.

Reading without ROS

The rosbags PyPI package reads ROS1 bags directly, no ROS install:

from huggingface_hub import snapshot_download
from rosbags.rosbag1 import Reader
from rosbags.typesys import Stores, get_typestore

local = snapshot_download("adipotnis/ts-polyline-bags",
                          repo_type="dataset",
                          allow_patterns="*.bag")
typestore = get_typestore(Stores.ROS1_NOETIC)
with Reader(f"{local}/ts_2023_04_10_20h05m59s_filtered.bag") as r:
    for conn, t, raw in r.messages():
        msg = typestore.deserialize_ros1(raw, conn.msgtype)
        ...

Related

• Processed dataset: adipotnis/ts-polyline (same traj_id keys)
• Build pipeline / viewer: adipotnis/ts-polyline-pipeline