gagandeepreehal/indus-ad-3d

Minus Zero Indian Urban Autonomous Driving Dataset - Multicamera LiDAR

Overview

This dataset provides original multicamera autonomous-driving recordings with LiDAR in MCAP format. It is designed for non-commercial research on camera and LiDAR perception, sensor synchronization, H.265 video pipelines, localization, GNSS/pose integration, and robotics data tooling.

The repository uses gated access for personal, educational, and research use under CC BY-NC 4.0. Commercial use is prohibited.

This focused raw release is part of the four-release Indian Urban Autonomous Driving collection. It is preserved without unnecessary clipping or fixed-rate subsampling so researchers can define their own camera–LiDAR synchronization, quality, and split policies.

Key facts

Property	Value
MCAP files	8
Total bytes	92,688,088,458
Approximate download size	92.69 GB
Total indexed camera duration	2:04:09.581
Total indexed LiDAR duration	2:06:43.895
Camera configuration	7 cameras reported
Camera identities observed	cam_front, cam_front_left, cam_front_right, cam_front_tilted, cam_rear, cam_rear_left, cam_rear_right
Container	MCAP
Camera payload	H.265/HEVC Annex-B access units
LiDAR schema	autonome.LidarFrame
Labels/annotations	None
Authored train/validation/test split	None; the manifest viewer's technical train label is not a benchmark split

Dataset structure

The original raw release contains .mcap recordings plus integrity and provenance control artifacts. The additive LiDAR companion below publishes only extracted LiDAR payloads and scan metadata. It does not publish MP4 derivatives, extracted camera frames, point-cloud annotations, or camera/GNSS copies.

README.md
manifest.jsonl
verification.json
data/
  <YYYY-MM-DD>/
    <recording>.mcap

Camera access units are stored in autonome.CompressedVideos messages and LiDAR records use autonome.LidarFrame. An H.265 access unit inside MCAP is not an MP4 file. The repository tree, MCAP summaries, embedded protobuf descriptors, and decoded values are authoritative; do not infer topology or timing from filenames alone.

manifest.jsonl remains the provenance manifest. data/tables/recordings/recordings.parquet is its equivalent eight-row Viewer index, allowing the recording and LiDAR-scan configurations to use one Parquet dataset builder.

Recording duration

Durations below sum each recording's indexed camera log-time span and preserve source timing. They are not wall-clock intervals between separate recordings.

Recording date	MCAP files	Camera duration	LiDAR duration
2025-10-22	4	2:02:03.389	2:04:37.396
2025-10-23	4	0:02:06.193	0:02:06.498
Total	8	2:04:09.581	2:06:43.895

Supplementary use with INDUS-AD

All eight LiDAR recordings correspond one-to-one with recordings in gagandeepreehal/indus-ad. Use the exact recording_id below as the cross-repository join key. The identifier is not translated or regenerated.

The correspondence was verified on 2026-10-25 against immutable revisions indus-ad-3d@ed567206 and indus-ad@59278055. For every listed recording, the exact ID is present in this repository's LiDAR extraction manifest, every row of its LiDAR scan table, and its LiDAR calibration row. The same exact ID is present in INDUS-AD's data/metadata.jsonl, in all seven video paths recorded there, and in every row of the linked cam_front camera-frame table.

Date	Shared recording_id	LiDAR scans	INDUS-AD cam_front frames	INDUS-AD camera metadata
2025-10-22	15-24-55.520-e317d15185a2a75e10e9a49b0f44db42	10,605	10,281	Parquet
2025-10-22	15-43-06.395-c746e0f897661ed0ba8799e1fe6b5aeb	28,576	28,327	Parquet
2025-10-22	16-40-30.395-84889bb142eb3a98aeacd10b5b26190d	6,746	5,065	Parquet
2025-10-22	17-08-42.892-62eb91c977f947a8b0c0e4eb9ac835a7	28,856	28,645	Parquet
2025-10-23	13-54-05.989-ef644bcba8d2e75a927420ae450726ee	214	204	Parquet
2025-10-23	13-54-47.414-cf95f0055261f746a994505e88d7c658	450	442	Parquet
2025-10-23	13-57-06.899-6e6941c5ebba4ee34c6c5fa4de393a47	375	374	Parquet
2025-10-23	13-58-22.898-26baac652baff76ab3b5eed15a059659	230	229	Parquet
Total	8 recordings	76,052	73,567

The shared identity metadata (recording_id and its SHA-256-derived two-character bucket) is identical across the two repositories. Modality-specific metadata is intentionally complementary rather than identical: INDUS-AD provides privacy-safe video, camera/GNSS, and annotation metadata, while this repository provides LiDAR payload, scan, and calibration metadata. Join on recording_id; do not join on the original MCAP filename.

The LiDAR scan table's time_offset_ns is expressed relative to the corresponding INDUS-AD recording's cam_front time origin. It supports temporal association, but timestamp proximity alone does not establish hardware synchronization or a physical camera-to-LiDAR transform.

Getting started

Downloading

python -m pip install -U huggingface_hub
REPO_ID="gagandeepreehal/minuszero-indian-autonomous-driving-multicam-lidar"
hf download "$REPO_ID" --repo-type dataset --include "*.mcap" --local-dir ./multicam-lidar-mcap

Download one recording from Python:

from huggingface_hub import hf_hub_download

mcap_path = hf_hub_download(
    repo_id="gagandeepreehal/minuszero-indian-autonomous-driving-multicam-lidar",
    filename="data/2025-10-22/2025-10-22_15-24-55.520.mcap",
    repo_type="dataset",
)
print(mcap_path)

Inspecting channels and messages

python -m pip install -U mcap mcap-protobuf-support

from mcap.reader import make_reader
from mcap_protobuf.decoder import DecoderFactory

path = "<path/to/recording.mcap>"
with open(path, "rb") as stream:
    reader = make_reader(stream, decoder_factories=[DecoderFactory()])
    for schema, channel, message, decoded in reader.iter_decoded_messages():
        print(schema.name, channel.topic, message.log_time, type(decoded).__name__)

Consumers should validate camera identities, array cardinality, timestamps, coordinate frames, point encoding, calibration, and codec parameter sets before use.

Channels and schema reference

Rows remain separate when the same protobuf name has a structurally different embedded descriptor. Files is unique-file coverage for that exact topic and layout; rows can overlap and must not be summed as a corpus total.

Topic	Protobuf schema	Descriptor	Files	Indexed messages
gnss	autonome.Pose	6a16b4479426	8	379,511
lidar	autonome.LidarFrame	4d4b9f718f6f	8	76,052
rec_cameras	autonome.CompressedVideos	53bc096b3b38	8	73,567
rec_gnss	autonome.Pose	6a16b4479426	8	73,567

How to read the protobuf tables

The tables below are generated from protobuf descriptors embedded in the MCAP summaries. Descriptor presence does not prove that every message populates a field. Repeated arrays can be empty or differ in length, and producer conventions govern unencoded units, coordinate frames, clock domains, calibration order, matrix layout, enums, and numeric ranges.

Complete protobuf schema layouts

autonome.CompressedVideos — descriptor 53bc096b3b38 (8 files)

Descriptor size: 1,473 bytes. Schema encoding: protobuf.

Descriptor files:

• google/protobuf/timestamp.proto — package google.protobuf, syntax proto3, dependencies: none
• Benchmarking.proto — package autonome, syntax proto3, dependencies: google/protobuf/timestamp.proto
• autonome/CompressedVideos.proto — package autonome, syntax proto3, dependencies: google/protobuf/timestamp.proto, Benchmarking.proto

google.protobuf.Timestamp

No.	Field	Cardinality	Protobuf type	Information available
1	seconds	optional	int64	Producer field seconds; units, range, and semantics are not further encoded by the int64 descriptor.
2	nanos	optional	int32	Producer field nanos; units, range, and semantics are not further encoded by the int32 descriptor.

autonome.Benchmarking

No.	Field	Cardinality	Protobuf type	Information available
1	sensor_source_time	optional	google.protobuf.Timestamp	Producer field sensor_source_time; units, range, and semantics are not further encoded by the google.protobuf.Timestamp descriptor.
2	process_start_time	optional	google.protobuf.Timestamp	Producer field process_start_time; units, range, and semantics are not further encoded by the google.protobuf.Timestamp descriptor.
3	data_receive_ack_time	repeated	google.protobuf.Timestamp	Producer field data_receive_ack_time; units, range, and semantics are not further encoded by the google.protobuf.Timestamp descriptor.
4	publish_time	optional	google.protobuf.Timestamp	Producer field publish_time; units, range, and semantics are not further encoded by the google.protobuf.Timestamp descriptor.
5	module_name	optional	string	Producer field module_name; units, range, and semantics are not further encoded by the string descriptor.
6	process_id	optional	uint64	Producer field process_id; units, range, and semantics are not further encoded by the uint64 descriptor.

autonome.CompressedVideos

No.	Field	Cardinality	Protobuf type	Information available
1	rec_frame_id	optional	int64	Producer field rec_frame_id; units, range, and semantics are not further encoded by the int64 descriptor.
2	rec_timestamp	optional	google.protobuf.Timestamp	Recorder timestamp.
3	timestamp	optional	google.protobuf.Timestamp	Message or sensor timestamp; its clock domain is producer-defined.
4	format	optional	string	Producer field format; units, range, and semantics are not further encoded by the string descriptor.
5	frame_id	optional	int64	Producer field frame_id; units, range, and semantics are not further encoded by the int64 descriptor.
6	data	repeated	bytes	Sensor payload bytes; interpret them using the enclosing schema and producer conventions.
7	name	repeated	string	Producer camera or sensor identity strings; validate array cardinality per message.
14	camera_timestamp	repeated	google.protobuf.Timestamp	Producer field camera_timestamp; units, range, and semantics are not further encoded by the google.protobuf.Timestamp descriptor.
8	width	optional	int32	Producer field width; units, range, and semantics are not further encoded by the int32 descriptor.
9	height	optional	int32	Producer field height; units, range, and semantics are not further encoded by the int32 descriptor.
10	number_of_cameras	optional	int32	Camera count reported by the producer for this message.
11	camera_intrinsic	repeated	autonome.CompressedVideos.CameraIntrinsic	Producer field camera_intrinsic; units, range, and semantics are not further encoded by the autonome.CompressedVideos.CameraIntrinsic descriptor.
12	camera_extrinsic	repeated	autonome.CompressedVideos.CameraExtrinsic	Producer field camera_extrinsic; units, range, and semantics are not further encoded by the autonome.CompressedVideos.CameraExtrinsic descriptor.
13	benchmarking	repeated	autonome.Benchmarking	Producer field benchmarking; units, range, and semantics are not further encoded by the autonome.Benchmarking descriptor.

autonome.CompressedVideos.CameraIntrinsic

No.	Field	Cardinality	Protobuf type	Information available
1	focal_length_x	optional	float	Producer field focal_length_x; units, range, and semantics are not further encoded by the float descriptor.
2	focal_length_y	optional	float	Producer field focal_length_y; units, range, and semantics are not further encoded by the float descriptor.
3	optical_center_x	optional	float	Producer field optical_center_x; units, range, and semantics are not further encoded by the float descriptor.
4	optical_center_y	optional	float	Producer field optical_center_y; units, range, and semantics are not further encoded by the float descriptor.
5	rmse	optional	float	Producer field rmse; units, range, and semantics are not further encoded by the float descriptor.
6	skew	optional	float	Producer field skew; units, range, and semantics are not further encoded by the float descriptor.
7	distortion_coeffs	repeated	double	Producer field distortion_coeffs; units, range, and semantics are not further encoded by the double descriptor.
8	width	optional	float	Producer field width; units, range, and semantics are not further encoded by the float descriptor.
9	height	optional	float	Producer field height; units, range, and semantics are not further encoded by the float descriptor.

autonome.CompressedVideos.CameraExtrinsic

No.	Field	Cardinality	Protobuf type	Information available
1	rotation_vector	repeated	float	Producer field rotation_vector; units, range, and semantics are not further encoded by the float descriptor.
2	translation_vector	repeated	float	Producer field translation_vector; units, range, and semantics are not further encoded by the float descriptor.

autonome.LidarFrame — descriptor 4d4b9f718f6f (8 files)

Descriptor size: 1,622 bytes. Schema encoding: protobuf.

Descriptor files:

• google/protobuf/timestamp.proto — package google.protobuf, syntax proto3, dependencies: none
• Benchmarking.proto — package autonome, syntax proto3, dependencies: google/protobuf/timestamp.proto
• LidarFrame.proto — package autonome, syntax proto3, dependencies: google/protobuf/timestamp.proto, Benchmarking.proto

google.protobuf.Timestamp

No.	Field	Cardinality	Protobuf type	Information available
1	seconds	optional	int64	Producer field seconds; units, range, and semantics are not further encoded by the int64 descriptor.
2	nanos	optional	int32	Producer field nanos; units, range, and semantics are not further encoded by the int32 descriptor.

autonome.Benchmarking

No.	Field	Cardinality	Protobuf type	Information available
1	sensor_source_time	optional	google.protobuf.Timestamp	Producer field sensor_source_time; units, range, and semantics are not further encoded by the google.protobuf.Timestamp descriptor.
2	process_start_time	optional	google.protobuf.Timestamp	Producer field process_start_time; units, range, and semantics are not further encoded by the google.protobuf.Timestamp descriptor.
3	data_receive_ack_time	repeated	google.protobuf.Timestamp	Producer field data_receive_ack_time; units, range, and semantics are not further encoded by the google.protobuf.Timestamp descriptor.
4	publish_time	optional	google.protobuf.Timestamp	Producer field publish_time; units, range, and semantics are not further encoded by the google.protobuf.Timestamp descriptor.
5	module_name	optional	string	Producer field module_name; units, range, and semantics are not further encoded by the string descriptor.
6	process_id	optional	uint64	Producer field process_id; units, range, and semantics are not further encoded by the uint64 descriptor.

autonome.LidarFrame

No.	Field	Cardinality	Protobuf type	Information available
1	frame_id	optional	int64	Producer field frame_id; units, range, and semantics are not further encoded by the int64 descriptor.
2	timestamp	optional	google.protobuf.Timestamp	Message or sensor timestamp; its clock domain is producer-defined.
3	column_timestamps	repeated	uint64	Producer field column_timestamps; units, range, and semantics are not further encoded by the uint64 descriptor.
10	columns_per_frame	optional	uint32	Producer field columns_per_frame; units, range, and semantics are not further encoded by the uint32 descriptor.
11	pixels_per_column	optional	uint32	Producer field pixels_per_column; units, range, and semantics are not further encoded by the uint32 descriptor.
12	rings	optional	uint32	Producer field rings; units, range, and semantics are not further encoded by the uint32 descriptor.
13	lidar_mode	optional	string	Producer field lidar_mode; units, range, and semantics are not further encoded by the string descriptor.
14	udp_profile_lidar	optional	string	Producer field udp_profile_lidar; units, range, and semantics are not further encoded by the string descriptor.
20	compression	optional	autonome.LidarFrame.Compression	Producer field compression; units, range, and semantics are not further encoded by the autonome.LidarFrame.Compression descriptor.
21	points_bin	optional	bytes	Producer field points_bin; units, range, and semantics are not further encoded by the bytes descriptor.
30	beam_intrinsics	optional	autonome.LidarFrame.BeamIntrinsics	Producer field beam_intrinsics; units, range, and semantics are not further encoded by the autonome.LidarFrame.BeamIntrinsics descriptor.
31	lidar_intrinsics	optional	autonome.LidarFrame.LidarIntrinsics	Producer field lidar_intrinsics; units, range, and semantics are not further encoded by the autonome.LidarFrame.LidarIntrinsics descriptor.
40	metadata_json	optional	string	Producer field metadata_json; units, range, and semantics are not further encoded by the string descriptor.
41	config_json	optional	string	Producer field config_json; units, range, and semantics are not further encoded by the string descriptor.
50	minimal_config	optional	autonome.LidarFrame.MinimalConfig	Producer field minimal_config; units, range, and semantics are not further encoded by the autonome.LidarFrame.MinimalConfig descriptor.
60	benchmarking	repeated	autonome.Benchmarking	Producer field benchmarking; units, range, and semantics are not further encoded by the autonome.Benchmarking descriptor.

Enum autonome.LidarFrame.Compression

Number	Value
0	NONE
1	ZSTD

autonome.LidarFrame.BeamIntrinsics

No.	Field	Cardinality	Protobuf type	Information available
1	beam_altitude_angles	repeated	float	Producer field beam_altitude_angles; units, range, and semantics are not further encoded by the float descriptor.
2	beam_azimuth_angles	repeated	float	Producer field beam_azimuth_angles; units, range, and semantics are not further encoded by the float descriptor.
3	beam_to_lidar_transform	repeated	float	Producer field beam_to_lidar_transform; units, range, and semantics are not further encoded by the float descriptor.
4	lidar_origin_to_beam_origin_mm	optional	float	Producer field lidar_origin_to_beam_origin_mm; units, range, and semantics are not further encoded by the float descriptor.

autonome.LidarFrame.LidarIntrinsics

No.	Field	Cardinality	Protobuf type	Information available
1	lidar_to_sensor_transform	repeated	float	Producer field lidar_to_sensor_transform; units, range, and semantics are not further encoded by the float descriptor.

autonome.LidarFrame.MinimalConfig

No.	Field	Cardinality	Protobuf type	Information available
1	timestamp_mode	optional	string	Producer field timestamp_mode; units, range, and semantics are not further encoded by the string descriptor.
2	udp_port_lidar	optional	uint32	Producer field udp_port_lidar; units, range, and semantics are not further encoded by the uint32 descriptor.
3	udp_port_imu	optional	uint32	Producer field udp_port_imu; units, range, and semantics are not further encoded by the uint32 descriptor.
4	udp_dest	optional	string	Producer field udp_dest; units, range, and semantics are not further encoded by the string descriptor.
5	azimuth_window	repeated	int32	Producer field azimuth_window; units, range, and semantics are not further encoded by the int32 descriptor.

autonome.Pose — descriptor 6a16b4479426 (8 files)

Descriptor size: 2,222 bytes. Schema encoding: protobuf.

Descriptor files:

• google/protobuf/timestamp.proto — package google.protobuf, syntax proto3, dependencies: none
• Pose.proto — package autonome, syntax proto3, dependencies: google/protobuf/timestamp.proto

google.protobuf.Timestamp

No.	Field	Cardinality	Protobuf type	Information available
1	seconds	optional	int64	Producer field seconds; units, range, and semantics are not further encoded by the int64 descriptor.
2	nanos	optional	int32	Producer field nanos; units, range, and semantics are not further encoded by the int32 descriptor.

autonome.Pose

No.	Field	Cardinality	Protobuf type	Information available
1	rec_frame_id	optional	int64	Producer field rec_frame_id; units, range, and semantics are not further encoded by the int64 descriptor.
2	rec_timestamp	optional	google.protobuf.Timestamp	Recorder timestamp.
3	timestamp	optional	google.protobuf.Timestamp	Message or sensor timestamp; its clock domain is producer-defined.
4	time_utc	optional	google.protobuf.Timestamp	UTC timestamp reported by the producer.
5	frame_id	optional	uint64	Producer field frame_id; units, range, and semantics are not further encoded by the uint64 descriptor.
6	position_type_info	optional	autonome.Pose.PositionTypeInfo	Producer field position_type_info; units, range, and semantics are not further encoded by the autonome.Pose.PositionTypeInfo descriptor.
7	lat_lon_ht	optional	autonome.Pose.LatLonHt	Producer field lat_lon_ht; units, range, and semantics are not further encoded by the autonome.Pose.LatLonHt descriptor.
8	enu	optional	autonome.Pose.ENU	Producer field enu; units, range, and semantics are not further encoded by the autonome.Pose.ENU descriptor.
9	local_lat_lon_ht	optional	autonome.Pose.LocalLatLonHt	Producer field local_lat_lon_ht; units, range, and semantics are not further encoded by the autonome.Pose.LocalLatLonHt descriptor.
10	local_enu	optional	autonome.Pose.LocalENU	Producer field local_enu; units, range, and semantics are not further encoded by the autonome.Pose.LocalENU descriptor.
11	velocity	optional	autonome.Pose.Velocity	Producer field velocity; units, range, and semantics are not further encoded by the autonome.Pose.Velocity descriptor.
12	orientation	optional	autonome.Pose.Orientation	Producer field orientation; units, range, and semantics are not further encoded by the autonome.Pose.Orientation descriptor.
13	orientation_error	optional	autonome.Pose.OrientationError	Producer field orientation_error; units, range, and semantics are not further encoded by the autonome.Pose.OrientationError descriptor.
14	position_error	optional	autonome.Pose.PositionError	Producer field position_error; units, range, and semantics are not further encoded by the autonome.Pose.PositionError descriptor.
15	vehicle_ref_pt	optional	autonome.Pose.VehicleReferencePoint	Producer field vehicle_ref_pt; units, range, and semantics are not further encoded by the autonome.Pose.VehicleReferencePoint descriptor.
16	is_valid	optional	bool	Producer field is_valid; units, range, and semantics are not further encoded by the bool descriptor.
17	re_init	optional	bool	Producer field re_init; units, range, and semantics are not further encoded by the bool descriptor.

Enum autonome.Pose.SpatialReferenceSystem

Number	Value
0	SRS_UNKNOWN
1	SRS_WEB_MERCATOR
2	SRS_WORLD_MERCATOR

Enum autonome.Pose.VehicleReferencePoint

Number	Value
0	REF_UNKNOWN
1	REF_ANTENNA_FRONT
2	REF_ANTENNA_REAR
3	REF_IMU
4	REF_WHEEL_AXLE_FRONT
5	REF_WHEEL_AXLE_REAR
6	REF_CAMERA_FRONT
7	REF_CAMERA_REAR
8	REF_LIDAR

autonome.Pose.PositionTypeInfo

No.	Field	Cardinality	Protobuf type	Information available
1	rtk_condition	optional	uint32	Producer field rtk_condition; units, range, and semantics are not further encoded by the uint32 descriptor.
2	correction_age_s	optional	float	Value in seconds.
3	position_fix_type	optional	uint32	Producer field position_fix_type; units, range, and semantics are not further encoded by the uint32 descriptor.

autonome.Pose.LatLonHt

No.	Field	Cardinality	Protobuf type	Information available
1	latitude_deg	optional	double	Angle or geographic coordinate in degrees.
2	longitude_deg	optional	double	Angle or geographic coordinate in degrees.
3	height_m	optional	double	Value in metres.

autonome.Pose.ENU

No.	Field	Cardinality	Protobuf type	Information available
1	east_m	optional	double	Value in metres.
2	north_m	optional	double	Value in metres.
3	up_m	optional	double	Value in metres.
4	ref_system_type	optional	autonome.Pose.SpatialReferenceSystem	Producer field ref_system_type; units, range, and semantics are not further encoded by the autonome.Pose.SpatialReferenceSystem descriptor.

autonome.Pose.LocalLatLonHt

No.	Field	Cardinality	Protobuf type	Information available
1	local_datum_id	optional	string	Producer field local_datum_id; units, range, and semantics are not further encoded by the string descriptor.
2	latitude_deg	optional	double	Angle or geographic coordinate in degrees.
3	longitude_deg	optional	double	Angle or geographic coordinate in degrees.
4	height_m	optional	double	Value in metres.

autonome.Pose.LocalENU

No.	Field	Cardinality	Protobuf type	Information available
1	local_east_m	optional	double	Value in metres.
2	local_north_m	optional	double	Value in metres.
3	local_datum_height_m	optional	double	Value in metres.

autonome.Pose.Velocity

No.	Field	Cardinality	Protobuf type	Information available
1	east_mps	optional	double	Value in metres per second.
2	north_mps	optional	double	Value in metres per second.
3	up_mps	optional	double	Value in metres per second.

autonome.Pose.Orientation

No.	Field	Cardinality	Protobuf type	Information available
1	roll_rad	optional	double	Angle in radians.
2	pitch_rad	optional	double	Angle in radians.
3	yaw_rad	optional	double	Angle in radians.

autonome.Pose.OrientationError

No.	Field	Cardinality	Protobuf type	Information available
1	pitch_variance_rad2	optional	double	Variance or squared angle in radians squared.
2	yaw_variance_rad2	optional	double	Variance or squared angle in radians squared.
3	roll_variance_rad2	optional	double	Variance or squared angle in radians squared.

autonome.Pose.PositionError

No.	Field	Cardinality	Protobuf type	Information available
1	position_rms	optional	double	Producer field position_rms; units, range, and semantics are not further encoded by the double descriptor.
2	east_sigma_m	optional	double	Value in metres.
3	north_sigma_m	optional	double	Value in metres.
4	up_sigma_m	optional	double	Value in metres.
5	covariance_en	optional	double	Producer field covariance_en; units, range, and semantics are not further encoded by the double descriptor.
6	pdop	optional	float	Producer field pdop; units, range, and semantics are not further encoded by the float descriptor.
7	hdop	optional	float	Producer field hdop; units, range, and semantics are not further encoded by the float descriptor.
8	vdop	optional	float	Producer field vdop; units, range, and semantics are not further encoded by the float descriptor.

Dataset Viewer

MCAP is a robotics log container, so embedded camera video and LiDAR are not natively previewable on the Hub. The default manifest.jsonl config indexes the eight original source MCAP files. The separate lidar_scans config indexes extracted LiDAR scans in Parquet; Viewer availability is subject to asynchronous processing. Download the MCAP files and use MCAP-aware tools locally.

Intended uses

Appropriate non-commercial uses include:

• multicamera and LiDAR perception or sensor-fusion research;
• robotics and MCAP reader development;
• H.265 extraction, indexing, synchronization, or remuxing experiments;
• LiDAR parsing, timing, calibration, and point-cloud tooling;
• non-commercial teaching, demonstrations, and personal projects.

The dataset is not validated for production driving systems, operational safety decisions, surveillance, identity recognition, or unsupported benchmark claims.

Limitations and known risks

• The dataset is unannotated and does not provide benchmark ground truth.
• Geographic, environmental, temporal, traffic, route, and hardware coverage may be uneven.
• Frames or LiDAR packets can be missing, delayed, duplicated, corrupted, or irregularly spaced.
• Camera and LiDAR streams can start or stop at different times; synchronize with source timestamps rather than array position or decoded index.
• Calibration and protobuf schema versions can differ between recordings.
• H.265 decoding support varies by platform and software build.
• Channel presence does not guarantee that every field is populated; validate each recording.
• LiDAR point encoding, units, axis ordering, handedness, coordinate frames, clock domains, and transforms are producer-defined unless the embedded descriptor states otherwise. Do not infer a hardware-trigger guarantee. See the sensor and timing conventions.

Privacy and responsible use

Recordings may have been captured in public-road environments and may incidentally contain people, faces, vehicle registration plates, residences, businesses, road signs, or precise GNSS/location traces. Do not assume that the data is anonymized.

Users are responsible for applicable privacy, data-protection, publicity, export-control, and local recording laws. Avoid re-identification, tracking individuals, exposing sensitive locations, or combining this dataset with external data in ways that increase privacy risk.

Do not post identifying evidence in a public discussion. Report sensitive, unlawful, or improperly included material privately through the Minus Zero contact form or by email to team@minuszero.ai, using the subject Dataset privacy or takedown request. Include the repository, recording_id or path when known, and enough information to locate the material. The recording-level removal policy applies across all four releases.

License

This dataset is released under the Creative Commons Attribution-NonCommercial 4.0 International license (CC BY-NC 4.0). Attribution is required and commercial use is prohibited. Sharing and adaptation are permitted for non-commercial purposes under the license terms. This card does not replace the license text, and other rights such as privacy, publicity, or trademark rights may still apply.

Citation

@dataset{minuszero_indian_autonomous_driving_multicam_lidar_2026,
  author    = {Gagandeep Reehal},
  title     = {Minus Zero Indian Urban Autonomous Driving Dataset - Multicamera LiDAR},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/gagandeepreehal/minuszero-indian-autonomous-driving-multicam-lidar}
}

Maintenance

Use repository discussions for non-sensitive dataset questions and corrections. Use the private contact route above for privacy and takedown reports.

Consumers should pin an immutable commit revision when reproducibility matters.

Extracted LiDAR companion

The original eight multicamera MCAP files remain the source records. This additive companion extracts only lidar topic messages into compressed point payloads and protobuf metadata; it contains no camera or GNSS payloads. Every scan has a Parquet index row under data/tables/lidar-scans/, and distinct LiDAR calibration/configuration values appear under data/tables/lidar-calibrations/. The payload archives are under data/lidar/; data/manifests/lidar-extraction.jsonl records paths, SHA-256 hashes, source revisions and counts. Each *.points.bin.zst archive member is the original ZSTD-compressed points_bin field. Its paired *.metadata.pb is the original LiDAR protobuf with points_bin cleared; restoring that field reconstructs the original message bytes. Scan indexes retain original sensor/log/publish timestamps and a nominal time offset relative to the matching INDUS-AD camera recording, using the documented UTC+05:30 conversion. The raw sensor timestamp is authoritative; the offset does not establish hardware synchronization. Point field semantics/units and camera-to-LiDAR projection accuracy are not independently validated. The original manifest.jsonl continues to index the eight source MCAP files.