shchon11/TCar

T-Car nuScenes — release 1002

Driving data from the T-Car test vehicle (six cameras, a roof LiDAR, RTK INS) in the nuScenes v1.0-trainval format: load it with the nuscenes-devkit as is.

268 scenes of 20 s (1.49 h) from 21 recordings · 10,720 key-frame samples · 991,534 sensor frames · 476 GB to download in 71 sensor tars + 1 table tar (555 GB unpacked). No 3D box annotations (sample_annotation is empty).

Download and use

1. Request access on this page (the dataset is gated), and log in: hf auth login.
2. Download this release and put it together (assemble.py checks every tar's sha256 and extracts it):

pip install -U huggingface_hub
hf download shchon11/TCar --repo-type dataset --revision 1002 --local-dir tcar_tars
python tcar_tars/assemble.py --tars tcar_tars --out tcar_nuscenes

from nuscenes.nuscenes import NuScenes
from nuscenes.can_bus.can_bus_api import NuScenesCanBus
nusc = NuScenes(version="v1.0-trainval", dataroot="tcar_nuscenes")
can = NuScenesCanBus(dataroot="tcar_nuscenes")      # pose, ms_imu, meta per scene

Some recordings only: download the tables and the recordings you want, and let assemble.py skip the rest (the tables still list every scene; the files of the others are just absent):

hf download shchon11/TCar --repo-type dataset --revision 1002 --local-dir tcar_tars \
    --include "manifest.json" --include "assemble.py" --include "meta/*" \
    --include "sensors/A-1_*/*" --include "sensors/B-2_*/*"
python tcar_tars/assemble.py --tars tcar_tars --out tcar_nuscenes --allow-missing

Updating: when a newer release is out, run the same two commands with its name as --revision, in the same directories: only the new tars are downloaded and extracted. Nothing already extracted is changed; the tables are replaced. Add --delete-tars to assemble.py to remove each tar after extracting it (then hf download fetches it again next time).

Releases and versioning

Each release is a branch of this repository (--revision <name>), a complete dataset on its own; a release is never changed after it is published. main always holds the latest release (a copy, stored once), so hf download without --revision gets it; give --revision to stay on one release. Later releases only add data:

• sensors/<recording>/<recording>.partNN.tar.gz — the sensor files (samples/, sweeps/) of a group of whole scenes of one recording, at most 10 GB unpacked (gzip: the LiDAR files shrink by about a third, the JPEGs not). Once published, a tar never changes and is carried into every later release (stored once).
• meta/TCar_meta.tar.gz — the tables (v1.0-trainval/), can_bus/, maps/, the per-recording *.import.json (how each recording was converted) and curation/. The only file that changes between releases.
• manifest.json — every tar with its recording, scene tokens, files, bytes and sha256, and the release history.
• Scene tokens never change. Scene names (scene-0001, …) are the nuScenes train split names in recording order; the names of released scenes do not change either. All scenes are in the train split.

release	date	recordings added	scenes added	scenes in total	parser commit
1002	2026-10-02	A-1, A-2, A-3, A-4, A-5, A-6, A-7, A-8, A-9, A-10, B-1, B-2, B-3, B-4, B-5, B-6, B-7, D-1, D-2, D-4, D-5	268	268	8a1f2f6

The 0923 branch holds the calibration sample recordings of 2026-09-23 (one tar per channel, its own scene numbering); they are not part of this series.

Contents

course	dates	recordings	scenes	time	GB
A	2026-09-28	10	119	40 min	224
B	2026-09-29	7	103	34 min	181
D	2026-09-29	4	46	15 min	71

channel	sensor	data	rate	frames
CAM_BACK	camera	1920×1080 JPEG	29 Hz	156,501
CAM_BACK_LEFT	camera	1920×1080 JPEG	29 Hz	156,501
CAM_BACK_RIGHT	camera	1920×1080 JPEG	29 Hz	156,501
CAM_FRONT	camera	1920×1080 JPEG	29 Hz	156,501
CAM_FRONT_LEFT	camera	1920×1080 JPEG	29 Hz	156,501
CAM_FRONT_RIGHT	camera	1920×1080 JPEG	29 Hz	156,501
LIDAR_TOP	lidar	point cloud (.pcd.bin)	10 Hz	52,528

Key frames (samples/) are 2 Hz, synchronised across the seven sensors as in nuScenes; every other frame is a sweep (sweeps/). LiDAR files hold five float32 per point (x, y, z, intensity, ring), as in nuScenes. calibrated_sensor holds the intrinsics and extrinsics, one set per recording.

Coordinate frames

ego_pose is in the location's global frame, as in nuScenes: metres, x east, y north, z up, on the tangent plane at 37.20° N, 126.83° E (height 0 on the WGS84 ellipsoid). log.json names it in the extra key global_frame (enu@37.200000,126.830000,0.000). Ego poses and the CAN bus pose come from the NovAtel INSPVA solution (100 Hz, RTK), placed at the receiver's GPS measurement time.

Recordings

recording	date	scenes	tars	GB	first release
A-1_2026-09-28-14-40-58	2026-09-28	10	3	18.5	1002
A-2_2026-09-28-14-45-11	2026-09-28	10	3	18.8	1002
A-3_2026-09-28-16-32-23	2026-09-28	12	3	22.4	1002
A-4_2026-09-28-14-51-38	2026-09-28	10	3	19.1	1002
A-5_2026-09-28-14-55-45	2026-09-28	12	3	23.1	1002
A-6_2026-09-28-15-02-17	2026-09-28	15	4	28.6	1002
A-7_2026-09-28-15-08-17	2026-09-28	12	3	21.7	1002
A-8_2026-09-28-15-13-01	2026-09-28	10	3	18.4	1002
A-9_2026-09-28-15-19-08	2026-09-28	12	3	23.1	1002
A-10_2026-09-28-16-03-14	2026-09-28	16	4	30.4	1002
B-1_2026-09-29-11-30-09	2026-09-29	12	3	21.5	1002
B-2_2026-09-29-11-37-49	2026-09-29	10	3	18.5	1002
B-3_2026-09-29-11-42-12	2026-09-29	17	5	31.1	1002
B-4_2026-09-29-11-50-32	2026-09-29	12	3	20.8	1002
B-5_2026-09-29-11-55-19	2026-09-29	13	4	22.5	1002
B-6_2026-09-29-12-02-16	2026-09-29	24	6	40.9	1002
B-7_2026-09-29-12-19-37	2026-09-29	15	4	25.8	1002
D-1_2026-09-29-17-08-46	2026-09-29	8	2	10.5	1002
D-2_2026-09-29-17-15-57	2026-09-29	13	3	20.6	1002
D-4_2026-09-29-17-33-25	2026-09-29	17	4	28.7	1002
D-5_2026-09-29-17-44-05	2026-09-29	8	2	10.7	1002

Maintainers: making the next release

Releases are made with scripts/hf_release.py of the parser repository (ROSbag_nuscenes_parser, branch tcar-gui; the commit of each release is in the table above):

# new recordings: parse them, curate them in the TCAR Parser app, press 최종 확정, then
python scripts/hf_release.py --branch <new> --from 1002 --token-file <hf token file>            # plan
python scripts/hf_release.py --branch <new> --from 1002 --token-file <hf token file> --upload   # build + upload

It refuses to publish unless every scene is locked and the dataset passes the checks (table references, devkit, CAN bus), and unless every tar already published still matches the dataset. Only the new recordings' tars are built and uploaded (one commit each, resumable); the tables, manifest, this card and assemble.py go last. First release of this series: 1002.