dmody1/race16-ros2-bag

Race16 ROS2 Bag Dataset

Extracted from ROS2 MCAP bag: race16_bag_20260118_120207.mcap

Contents

Images (data/images/)

• 21 shards of color (JPEG) + depth (16-bit PNG) image pairs
• ~10,155 color frames, ~10,156 depth frames
• 500 image pairs per shard (tar.gz compressed)
• ~30 FPS, ~339 seconds duration

Proprioceptive Data (data/)

File	Rows	Description
imu.csv	67,410	Orientation, angular velocity, linear acceleration
odometry.csv	36,349	Position, orientation, linear/angular velocity
position_cmd.csv	36,335	Position, velocity, acceleration, jerk, yaw commands
so3_cmd.csv	36,331	Force, orientation, angular velocity, controller gains
image_index.csv	10,155	Frame ID to file path + timestamp mapping

All CSVs use timestamp_ns (nanoseconds) as the first column for synchronization.

Loading

import pandas as pd
from huggingface_hub import hf_hub_download, snapshot_download

# Download CSVs
snapshot = snapshot_download(repo_id="dmody1/race16-ros2-bag", repo_type="dataset")

imu = pd.read_csv(f"{{snapshot}}/data/imu.csv")
odom = pd.read_csv(f"{{snapshot}}/data/odometry.csv")

# Extract image shards
import tarfile, os
shard_path = f"{{snapshot}}/data/images/shard_0000.tar.gz"
with tarfile.open(shard_path, "r:gz") as tar:
    tar.extractall("images/")