Datasets:

nvidia
/

Cosmos3-Driving-Simulation-Evaluation

sample_id stringclasses 91 values	camera_key stringclasses 11 values	caption_path stringlengths 46 51	caption stringlengths 1.64k 5.49k	width int64 1.35k 1.92k	height int64 1.08k 1.08k	fps float64 30 30	num_frames int64 351 657	duration_seconds float64 11.7 21.9	caption_end_frame int64 300 300
scene-001	camera_front_wide_120fov	captions/scene-001/camera_front_wide_120fov.json	[0.0s-4.5s]: The view is a first-person forward-facing perspective from inside a moving vehicle, with the light-colored hood of a Mercedes-Benz (bearing the brand's three-pointed star emblem at its center) visible in the immediate foreground. The vehicle travels straight along a smooth paved multi-lane highway marked w...	1,924	1,084	30	654	21.8	300
scene-001	camera_cross_left_120fov	captions/scene-001/camera_cross_left_120fov.json	[0s-5.4s]: A first-person forward-facing view from a moving vehicle travels along a smooth dark asphalt multi-lane highway, with the camera translating steadily forward at eye level in wide landscape framing. The foreground right lane is marked by a solid orange edge line and white dashed lane dividers. A continuous se...	1,924	1,084	30	654	21.8	300
scene-001	camera_cross_right_120fov	captions/scene-001/camera_cross_right_120fov.json	[0.0s-5.0s]: The view is a first-person perspective from a moving vehicle traveling forward along a smooth dark asphalt road, with a solid orange edge line running along the left portion of the visible pavement. A continuous silver metal W-beam guardrail supported by evenly spaced vertical posts runs parallel to the ro...	1,924	1,084	30	654	21.8	300
scene-001	camera_rear_left_70fov	captions/scene-001/camera_rear_left_70fov.json	[0.0s-5.6s]: The view is a first-person eye-level perspective from a vehicle traveling forward on a multi-lane asphalt highway under overcast, gray cloudy skies. Steep mountains with snow-capped upper peaks and mixed forested lower slopes rise in the background; tall metal electricity pylons stand along the mountain ba...	1,924	1,084	30	654	21.8	300
scene-001	camera_rear_right_70fov	captions/scene-001/camera_rear_right_70fov.json	[0.0s-5.0s]: The view is a first-person eye-level perspective from a moving vehicle traveling forward along a smooth dark asphalt road, with the partial side of a light-colored vehicle visible at the far right edge of the frame for the full duration. A continuous silver W-beam metal guardrail supported by evenly spaced...	1,924	1,084	30	654	21.8	300
scene-001	camera_rear_tele_30fov	captions/scene-001/camera_rear_tele_30fov.json	[0.0s-3.0s]: A forward-facing, eye-level view captures travel along a multi-lane asphalt highway under overcast, hazy conditions, with distant snow-dusted mountain peaks visible beneath heavy gray clouds. Steep mountainsides covered in mixed evergreen and bare deciduous trees rise on both sides of the road; tall lattic...	1,924	1,084	30	654	21.8	300
scene-001	camera_front_tele_30fov	captions/scene-001/camera_front_tele_30fov.json	[0.0s-4.0s]: The view is a first-person forward-facing perspective from a vehicle traveling along a straight, multi-lane asphalt highway in a mountain valley. Two solid red lines mark the left and right edges of the current travel lane; white dashed lane markings separate adjacent same-direction lanes to the left, and ...	1,924	1,084	30	654	21.8	300
scene-001	camera_front_fisheye_200fov	captions/scene-001/camera_front_fisheye_200fov.json	[0.0s-5.5s]: The view is a first-person forward-facing perspective from a moving vehicle, captured with a fisheye lens that creates dark circular vignetting around the frame edges and curved distortion of straight lines. The vehicle travels along a smooth gray asphalt highway lane marked by solid orange lines on both i...	1,720	1,080	30	654	21.8	300
scene-001	camera_left_fisheye_200fov	captions/scene-001/camera_left_fisheye_200fov.json	[0.0s-4.8s]: The view is a first-person perspective from a moving vehicle, captured with a fisheye lens that creates curved distortion of road lines and the horizon, with dark circular vignetting around the frame edges; portions of the white host vehicle's body are visible at the bottom left and bottom right corners. T...	1,720	1,080	30	654	21.8	300
scene-001	camera_rear_fisheye_200fov	captions/scene-001/camera_rear_fisheye_200fov.json	[0.0s-5.6s]: The view is a first-person forward-facing perspective from inside a moving vehicle, framed by the curved white interior edge of the windshield opening with dark vignetting around the circular field of view; partial black stylized text is visible on the lower left interior frame. The setting is a straight m...	1,720	1,080	30	654	21.8	300
scene-001	camera_right_fisheye_200fov	captions/scene-001/camera_right_fisheye_200fov.json	[0.0s-5.0s]: The view is a first-person forward-facing perspective from a moving vehicle, captured with a fisheye lens that creates strong barrel distortion and dark circular vignetting around the frame edges. The lower left and right of the frame show portions of the host vehicle's light-colored body panels; a sticker...	1,720	1,080	30	654	21.8	300
scene-002	camera_front_wide_120fov	captions/scene-002/camera_front_wide_120fov.json	[0.0s-1.5s]: The view is a first-person dashcam perspective from inside a Mercedes-Benz vehicle, with the light-colored hood and centered Mercedes three-pointed star logo visible in the immediate foreground, traveling forward on a damp multi-lane highway under an overcast gray sky. Directly ahead in the right travel la...	1,924	1,084	30	654	21.8	300
scene-002	camera_cross_left_120fov	captions/scene-002/camera_cross_left_120fov.json	[0.0s-2.1s]: The view is a first-person forward-facing perspective from a vehicle traveling on a dark asphalt multi-lane highway under uniformly overcast, gray, flat lighting with no harsh shadows. A tall gray concrete median barrier runs along the left side of the travel lanes, separating opposing traffic; a white rib...	1,924	1,084	30	654	21.8	300
scene-002	camera_cross_right_120fov	captions/scene-002/camera_cross_right_120fov.json	[0.0s-1.0s]: The view is a first-person perspective from a moving vehicle traveling along an asphalt road, with a white zigzag chevron road marking running along the left edge of the pavement in the immediate foreground. A curved grey metal W-beam guardrail supported by vertical posts lines the road edge, separating it...	1,924	1,084	30	654	21.8	300
scene-002	camera_rear_left_70fov	captions/scene-002/camera_rear_left_70fov.json	[0.0s-1.5s]: The view is a first-person forward-facing perspective from a vehicle traveling on a wet, dark asphalt multi-lane road under uniformly overcast gray skies. A concrete Jersey barrier runs along the right edge of the travel lanes, separating opposing traffic directions; white dashed lines mark lane divisions,...	1,924	1,084	30	654	21.8	300
scene-002	camera_rear_right_70fov	captions/scene-002/camera_rear_right_70fov.json	[0.0s-2.5s]: The view is a first-person eye-level perspective from a moving vehicle traveling forward along a paved asphalt road, with the edge of the filming vehicle's light-colored body visible at the far right of the frame. A grey metal W-beam guardrail runs parallel to the road in the immediate left foreground, sep...	1,924	1,084	30	654	21.8	300
scene-002	camera_rear_tele_30fov	captions/scene-002/camera_rear_tele_30fov.json	[0.0s-4.8s]: The scene takes place on a multi-lane divided asphalt road under overcast, hazy gray conditions with muted low contrast and soft diffused lighting; all visible vehicles have their headlights illuminated. The camera travels forward at eye level in a first-person driving perspective, framing the two travel l...	1,924	1,084	30	654	21.8	300
scene-002	camera_front_tele_30fov	captions/scene-002/camera_front_tele_30fov.json	[0.0s-2.0s]: The view is a first-person forward-facing perspective from a vehicle traveling on a paved multi-lane road under overcast, flat gray lighting with muted color saturation. Directly ahead in the right lane sits a large blue box truck with visible scuffs and wear on its rear panel, illuminated red taillights, ...	1,924	1,084	30	654	21.8	300
scene-002	camera_front_fisheye_200fov	captions/scene-002/camera_front_fisheye_200fov.json	[0.0s-2.6s]: The view is a first-person forward-facing dashcam perspective with strong fisheye lens distortion and dark circular vignetting around the frame edges, capturing travel along a wet multi-lane asphalt highway under overcast, gray cloudy skies. A teal box truck travels directly ahead in the same lane as the c...	1,720	1,080	30	654	21.8	300
scene-002	camera_left_fisheye_200fov	captions/scene-002/camera_left_fisheye_200fov.json	[0.0s-2.5s]: The view is a first-person forward-facing perspective from a moving vehicle, captured with a fisheye lens that creates curved distortion at the frame edges; the light-colored hood of the camera vehicle is visible along the bottom edge of the frame. The scene takes place on a multi-lane divided asphalt high...	1,720	1,080	30	654	21.8	300
scene-002	camera_rear_fisheye_200fov	captions/scene-002/camera_rear_fisheye_200fov.json	[0.0s-1.5s]: The view is a first-person forward-facing perspective from a moving vehicle, captured with a fisheye lens that creates curved edge distortion and a dark circular vignette around the frame; portions of the host vehicle's white body are visible at the bottom left and right edges of the frame. The setting is ...	1,720	1,080	30	654	21.8	300
scene-002	camera_right_fisheye_200fov	captions/scene-002/camera_right_fisheye_200fov.json	[0.0s-2.5s]: The view is a first-person forward perspective from inside a moving vehicle, captured with a fisheye lens that creates strong curved distortion at the frame edges; white interior car trim and a partial dark sticker with unreadable text are visible along the bottom left and right borders of the frame. A sil...	1,720	1,080	30	654	21.8	300
scene-003	camera_front_wide_120fov	captions/scene-003/camera_front_wide_120fov.json	[0.0s-4.5s]: The view is a first-person forward perspective from inside a moving vehicle, with the vehicle's hood visible in the lower foreground and a small black Mercedes-Benz three-pointed star emblem centered on the hood. The scene takes place at night under a dark sky, with slight haze creating soft glare around a...	1,924	1,084	30	654	21.8	300
scene-003	camera_cross_left_120fov	captions/scene-003/camera_cross_left_120fov.json	[0s-2.0s]: The scene is a nighttime multi-lane asphalt road, captured from a first-person forward-moving perspective consistent with dashcam footage. Double yellow lines run down the center of the travel lanes, and a solid white line marks the right edge of the lane. Tall gray metal streetlight poles with illuminated c...	1,924	1,084	30	654	21.8	300
scene-003	camera_cross_right_120fov	captions/scene-003/camera_cross_right_120fov.json	[0s-3.5s]: The view is a first-person perspective from a moving vehicle traveling forward on a paved multi-lane urban road at night. The dark asphalt road has white dashed lane markings, and a concrete curb runs along the right edge separating the road from a sidewalk lined with dense, dark green leafy trees. Behind th...	1,924	1,084	30	654	21.8	300
scene-003	camera_rear_left_70fov	captions/scene-003/camera_rear_left_70fov.json	[0s-3.5s]: The view is a first-person forward perspective from a moving vehicle traveling along a paved multi-lane urban road at night, with the left edge of the filming vehicle's light-colored body visible along the far left of the frame throughout the sequence. The road features double yellow center lines and white l...	1,924	1,084	30	654	21.8	300
scene-003	camera_rear_right_70fov	captions/scene-003/camera_rear_right_70fov.json	[0.0s-3.5s]: The view is a first-person perspective from a moving vehicle traveling forward on a multi-lane asphalt road at night, with the edge of the filming vehicle's white body visible along the far right edge of the frame for the full duration. White dashed and solid lane markings divide the road lanes. Oncoming t...	1,924	1,084	30	654	21.8	300
scene-003	camera_rear_tele_30fov	captions/scene-003/camera_rear_tele_30fov.json	[0.0s-5.0s]: The view is a first-person forward-facing perspective from a moving vehicle traveling at night on a straight multi-lane paved asphalt road, with slight haze and prominent vertical lens flare streaks cast by bright light sources across the frame. In the left adjacent lane, a dark-colored SUV travels in the ...	1,924	1,084	30	654	21.8	300
scene-003	camera_front_tele_30fov	captions/scene-003/camera_front_tele_30fov.json	[0.0s-3.5s]: The view is a first-person forward-facing perspective from a moving vehicle at a nighttime signalized intersection, with slight haze and lens flare from bright lights reducing long-distance visibility. Overhead horizontal traffic signal poles span the road: green street signs with white text reading "Ringw...	1,924	1,084	30	654	21.8	300
scene-003	camera_front_fisheye_200fov	captions/scene-003/camera_front_fisheye_200fov.json	[0s-3.5s]: First-person forward-facing dashcam view with slight fisheye distortion and dark circular vignetting from the vehicle's windshield frame, traveling straight on a multi-lane asphalt road at night. The road is marked with white lane lines: a solid white line runs along the left edge of the travel lanes separat...	1,720	1,080	30	654	21.8	300
scene-003	camera_left_fisheye_200fov	captions/scene-003/camera_left_fisheye_200fov.json	[0.0s-2.5s]: The view is a first-person perspective from a moving white vehicle, captured with a fisheye lens that creates curved edge distortion and visible light glare from bright sources. The scene takes place at night on a multi-lane paved asphalt road marked with white lane lines and a painted left-turn arrow in t...	1,720	1,080	30	654	21.8	300
scene-003	camera_rear_fisheye_200fov	captions/scene-003/camera_rear_fisheye_200fov.json	[0s-3.5s]: The view is a first-person perspective from inside a moving vehicle, framed by the white lower edge of the windshield at the bottom of the frame and a circular fisheye lens distortion bordering the entire view. It is nighttime, with a dark sky overhead. A multi-lane paved asphalt road stretches straight ahea...	1,720	1,080	30	654	21.8	300
scene-003	camera_right_fisheye_200fov	captions/scene-003/camera_right_fisheye_200fov.json	[0.0s-3.5s]: First-person forward-facing view from a moving vehicle, captured with a wide-angle fisheye lens that produces curved edge distortion and prominent starburst lens flares around bright light sources. The scene takes place at night on a dark asphalt multi-lane road marked with white dashed lane lines; the low...	1,720	1,080	30	654	21.8	300
scene-004	camera_front_wide_120fov	captions/scene-004/camera_front_wide_120fov.json	[0.0s-3.5s]: The view is a first-person forward-facing perspective from inside a moving vehicle, with the Mercedes-Benz three-pointed star logo visible on the hood at bottom center, and slight wide-angle fisheye distortion curving frame edges. The scene is a sunlit urban street under a clear, cloudless bright blue sky,...	1,920	1,080	30	651	21.7	300
scene-004	camera_cross_left_120fov	captions/scene-004/camera_cross_left_120fov.json	[0.0s-3.5s]: The scene is a sunlit urban street under a clear, cloudless deep blue sky, with warm low-angle sunlight casting long, sharp dark shadows across the road and building facades. A tall, curved light beige high-rise building dominates the midground center, with a large shadow of an adjacent structure cast on i...	1,920	1,080	30	651	21.7	300
scene-004	camera_cross_right_120fov	captions/scene-004/camera_cross_right_120fov.json	[0.0s-2.1s]: A silver four-door sedan travels from frame left to frame right across the foreground, passing close to the viewpoint before exiting frame right. Parked along the left curb behind it are a dark SUV and a smaller light-colored passenger car. The background is dominated by a multi-story red brick building wi...	1,920	1,080	30	651	21.7	300
scene-004	camera_rear_left_70fov	captions/scene-004/camera_rear_left_70fov.json	[0.0s-2.5s]: The view is a wide, eye-level first-person perspective from a moving vehicle traveling forward along an urban asphalt road, with the partial side of a white car visible in the immediate left foreground throughout. A raised concrete traffic island sits in the near midground, holding two circular blue-border...	1,920	1,080	30	651	21.7	300
scene-004	camera_rear_right_70fov	captions/scene-004/camera_rear_right_70fov.json	[0.0s-2.5s]: The view is a first-person eye-level perspective from a moving vehicle traveling backward along a narrow urban street, with the front edge of a light-colored parked car visible along the far right edge of the frame for the full duration. Multi-story European-style buildings line both sides of the dark asph...	1,920	1,080	30	651	21.7	300
scene-004	camera_rear_tele_30fov	captions/scene-004/camera_rear_tele_30fov.json	[0.0s-3.5s]: A wide, eye-level view captures a busy urban street in cool late-afternoon light, with the camera moving steadily backward down the road center. The asphalt road has visible wet patches and repair marks, particularly on the right side. In the center lane, a white Renault van with yellow license plate faces...	1,920	1,080	30	651	21.7	300
scene-004	camera_front_tele_30fov	captions/scene-004/camera_front_tele_30fov.json	[0.0s-3.5s]: A wide, eye-level view moves steadily forward along an urban asphalt street in bright daytime sunlight, with low-angle warm light casting long shadows and gilding upper building facades. Parked cars line both curbs: on the left, a white Ford van, white Renault hatchback, silver sedan, and dark-colored cars...	1,920	1,080	30	651	21.7	300
scene-004	camera_front_fisheye_200fov	captions/scene-004/camera_front_fisheye_200fov.json	[0.0s-3.5s]: The view is a first-person perspective from a moving vehicle traveling straight forward on a paved urban street, captured with a fisheye lens that creates curved distortion at the frame edges and dark vignetting around the circular field of view. Bright clear daylight illuminates the scene under a cloudles...	1,350	1,080	30	651	21.7	300
scene-004	camera_left_fisheye_200fov	captions/scene-004/camera_left_fisheye_200fov.json	[0.0s-7.0s]: The view is a first-person fisheye perspective from a moving white vehicle, with the vehicle's white hood visible in the immediate foreground; a black sticker with reversed white text reading "Test Vehicle - Volvo" and a small logo is affixed to the lower left portion of the hood. The scene is a busy urban...	1,350	1,080	30	651	21.7	300
scene-004	camera_rear_fisheye_200fov	captions/scene-004/camera_rear_fisheye_200fov.json	[0.0s-3.5s]: The view is a first-person forward-facing perspective from a moving motorcycle, captured with a fisheye lens that creates curved distortion at the frame edges. A white motorcycle fairing occupies the immediate foreground, marked with large black stylized graphics and a blue EU license plate strip on its ri...	1,350	1,080	30	651	21.7	300
scene-004	camera_right_fisheye_200fov	captions/scene-004/camera_right_fisheye_200fov.json	[0.0s-4.5s]: The view is a first-person perspective from a moving vehicle, captured with a wide fisheye lens that creates curved distortion at the frame edges and a dark circular vignette border; the white hood of the filming vehicle occupies the bottom foreground, with a black sticker bearing reversed white text "Team...	1,350	1,080	30	651	21.7	300
scene-005	camera_front_wide_120fov	captions/scene-005/camera_front_wide_120fov.json	[0.0s-4.0s]: The view is a first-person forward-facing perspective from inside a moving vehicle, with the top edge of the vehicle's dashboard visible at the bottom of the frame showing small reflected indicator lights, traveling on a paved multi-lane urban road at night. The dark sky has soft glare and lens flare from ...	1,920	1,080	30	651	21.7	300
scene-005	camera_cross_left_120fov	captions/scene-005/camera_cross_left_120fov.json	[0.0s-3.5s]: The scene is a nighttime urban road intersection, viewed from a first-person forward-moving perspective at eye level in wide framing. The foreground consists of smooth dark asphalt marked with crisp white painted lane lines, including directional turn arrows and crosswalk markings. A grassy median runs hor...	1,920	1,080	30	651	21.7	300
scene-005	camera_cross_right_120fov	captions/scene-005/camera_cross_right_120fov.json	[0.0s-1.5s]: The scene is a nighttime urban street corner, captured from a first-person eye-level perspective as the camera moves forward along the road. In the right foreground, a curved rough-textured gray concrete planter holds a large leafless deciduous tree with pale bare branches; dry brown leaves are scattered o...	1,920	1,080	30	651	21.7	300
scene-005	camera_rear_left_70fov	captions/scene-005/camera_rear_left_70fov.json	[0.0s-3.5s]: The view is a first-person, eye-level wide shot from a vehicle traveling forward along a paved urban road at night. The left edge of the frame is partially occupied by the side of a white car; immediately behind it in the same left lane, a dark-colored car with illuminated bright white headlights faces tow...	1,920	1,080	30	651	21.7	300
scene-005	camera_rear_right_70fov	captions/scene-005/camera_rear_right_70fov.json	[0.0s-2.5s]: The scene is a nighttime urban street, captured from a first-person eye-level perspective moving steadily forward along the road. In the immediate right foreground, the side of a white car is partially visible, with a dark-colored car positioned directly ahead of it in the same travel lane; both vehicles h...	1,920	1,080	30	651	21.7	300
scene-005	camera_rear_tele_30fov	captions/scene-005/camera_rear_tele_30fov.json	[0.0s-2.5s]: A black Audi sedan with illuminated bright white LED headlights sits in the center foreground, facing directly toward the viewer on a nighttime urban road. The driver is visible through the front windshield, holding one hand up to their face near their mouth. The car's front license plate is partially obsc...	1,920	1,080	30	651	21.7	300
scene-005	camera_front_tele_30fov	captions/scene-005/camera_front_tele_30fov.json	[0.0s-8.5s]: The scene is a nighttime urban intersection, viewed from a first-person eye-level perspective in a moving vehicle traveling forward along the road. A blue and silver public transit bus with illuminated red taillights and a yellow digital route display on its rear occupies the center lane directly ahead, dr...	1,920	1,080	30	651	21.7	300
scene-005	camera_front_fisheye_200fov	captions/scene-005/camera_front_fisheye_200fov.json	[0.0s-3.5s]: The view is a first-person forward-facing perspective from a moving vehicle, captured with a fisheye lens that creates curved distortion at the frame edges; the lower edge of the frame shows the dark curved hood and light-colored front bumper trim of the vehicle. The scene takes place at night on a multi-l...	1,350	1,080	30	651	21.7	300

End of preview. Expand in Data Studio

Cosmos3 Driving Simulation Evaluation

Dataset Summary

This dataset contains 91 driving scenes with 11 camera views per scene: 1,001 World Scenario Map (WSM) control videos and 1,001 matching timestamped caption JSON files.

Dataset Description

The dataset provides structured scene controls and text prompts for multiview driving generation and evaluation. The WSM controls are derived from real-world driving data collected by NVIDIA and represent lane lines, object bounding boxes, and traffic-light states.

Included: WSM 4.0.1 rendered control videos, per-camera captions, scene/view manifests, measured video properties, and file checksums.

Not included: source RGB camera images or videos, LiDAR data, calibration, raw scene annotations, or model-generated outputs. WSM videos are abstract renderings of scene structure, not camera recordings. Original or restyled first-frame images are not included.

The released manifest identifies no separate standard or long-rollout subsets; control videos are 11.7–21.9 seconds long.

Dataset Owner(s)

NVIDIA Corporation

Dataset Creation Date

2026-10-20

Versioning

V1.0

No previous version.

License/Terms of Use

This dataset is released under the OpenMDW License Agreement, version 1.1 (OpenMDW-1.1).

Intended Usage

This dataset is intended for autonomous vehicle-related use cases only, including both commercial and non-commercial use in accordance with the applicable license terms. Its size and distribution across camera views support synthetic driving scene evaluation.

Public text-to-video evaluation of Cosmos driving simulation using WSM controls and accompanying text prompts. Specifically:

evaluating Cosmos3 driving simulation conditioned on WSM controls.

This dataset is not intended for training perception or driving models.

Dataset Characterization

Data Collection Method: Automatic/Sensors.

Sensor-collected driving data is processed into WSM controls.

Labeling Method: Automated.

Human reviewers provide quality assurance for the automated annotations. Processing steps:

NVIDIA in-house models annotate lane lines and object bounding boxes on the source footage.
Human reviewers check the annotations for quality.
A script converts the reviewed annotations into WSM controls.
An internal NVIDIA caption model generates a text caption per camera view, used as a text prompt for evaluation.
All camera imagery is discarded. Only the WSM controls and text prompts are released.

Dataset Format

Modality: Video and Text.

WSM control videos in mp4 format, at 30fps. Each video shows the scene layout over time: lane lines, object bounding boxes, and traffic-light states.

Text captions are provided as 11 separate JSON files per scene, one per camera view.

Dataset Quantification

Record count: 91 multiview sequences (11 camera views per sequence).

Size bin: n<1K, where each record is one sequence.

Video length: 351–657 frames (11.7–21.9 seconds) at 30fps.

Resolution:

Seven standard cameras: 1924 × 1084 for 53 sequences, or 1920 × 1080 for 38 sequences.
Four fisheye cameras: 1720 × 1080 for 53 sequences, or 1350 × 1080 for 38 sequences.

Total storage size: approximately 2.59 GB (decimal).

Limitations

Small sample. 91 scenes cover a limited range of road types, regions, traffic densities, and weather. Results on this evaluation subset may not generalize to other driving scenarios.
One conditioning format. Results apply to WSM conditioning and may not transfer to other conditioning formats, such as depth, segmentation, or edge maps.
Annotation error. Lane lines and bounding boxes come from automated models with human QA. Some missed or misplaced objects may remain.
Layout only. WSM encodes scene geometry and traffic-light state. It does not encode appearance, lighting, weather, or road-surface detail.

Reference(s)

WSM representation example. The top half shows WSM controls; the bottom half shows Cosmos-generated video. The released dataset contains WSM controls and text prompts, without generated video.

Ethical Considerations

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. Developers should work with their internal developer teams to ensure this dataset meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Privacy: The dataset contains no camera images, faces, license plates, or legible text. Road agents appear only as bounding boxes.

Please report quality, risk, security vulnerabilities or NVIDIA AI Concerns here.

Files

wsm/scene-001/camera_front_wide_120fov.mp4
captions/scene-001/camera_front_wide_120fov.json
...
manifest.jsonl
metadata.jsonl
checksums.json

manifest.jsonl: one record per scene, grouping all 11 views by camera_key, with repository-relative wsm_path and caption_path values.
metadata.jsonl: one record per camera view, linking its control video to the complete caption text and measured video properties for the dataset viewer.
checksums.json: SHA-256 and byte size for every other release payload file.
scene-001 through scene-091 are release-local identifiers.

The viewer has 1,001 rows, not 1,001 independent scenes. Group rows by sample_id to recover the 91 multiview scenes. The test split contains the entire evaluation collection; no training split is supplied.

Browse the dataset

Use the built-in Dataset Viewer / Data Studio on this dataset page, with configuration default and split test. Each row contains one playable WSM video and its matching timestamped caption.

Search for a scene ID such as scene-001 to find its 11 camera views.
Use camera_key to identify the view, for example camera_front_wide_120fov.
Read caption for the complete prompt. caption_path points to the matching structured JSON in the Files tab.
Check width, height, fps, num_frames, and duration_seconds for the native control-video properties. caption_start_frame and caption_end_frame describe the caption window, not the full video duration.
Download a complete scene using the command in the Download section below.

The built-in viewer displays camera videos individually; it is not a synchronized surround player. It shows only WSM controls and captions, not RGB recordings or LiDAR. Hosted browsing requires the data to be on the dataset's default branch and processed by Hugging Face; files proposed only in a pull request are not a verified live viewer. Restricted repositories require access.

Camera names

Each scene contains these camera keys. Match files by camera key, not directory listing order.

camera_front_wide_120fov
camera_cross_left_120fov
camera_cross_right_120fov
camera_rear_left_70fov
camera_rear_right_70fov
camera_rear_tele_30fov
camera_front_tele_30fov
camera_front_fisheye_200fov
camera_left_fisheye_200fov
camera_rear_fisheye_200fov
camera_right_fisheye_200fov

Captions and timing

Each JSON preserves the structured caption format:

{
  "caption_structured": {
    "chunk_0_300": {
      "caption": "[0.0s-2.0s]: ...",
      "start_frame": 0,
      "end_frame": 300
    }
  }
}

This is a schema illustration, not an additional sample. Use each file's actual chunk keys, frame bounds and timestamped text.

The captions are the timestamped Tier1 version, with a privacy cleanup of identifying text. They are not a new environmental prompt rewrite. The cleanup removes transcribed identifying values, including license-plate values, while retaining generic descriptions such as a vehicle having a visible plate. Timestamp strings, frame bounds and other JSON fields are preserved.

Control videos retain their original bytes, frame rate, dimensions and duration. A control video may be longer than its caption window. Use the caption's frame bounds when selecting a generation interval; do not assume that a caption describes the entire MP4. metadata.jsonl records measured control dimensions, frame count, frame rate and duration. The upload does not crop, retime, resize or re-encode the controls.

Captions are machine-generated descriptions and may contain errors. Privacy cleanup is not a correction of all scene descriptions or an independent verification of physical camera calibration.

Download

Download the files directly; there is no ZIP extraction step:

hf download nvidia/Cosmos3-Driving-Simulation-Evaluation \
  --repo-type dataset --local-dir driving-simulation-evaluation

For one scene, download its two folders plus the manifests:

hf download nvidia/Cosmos3-Driving-Simulation-Evaluation \
  --repo-type dataset --local-dir driving-simulation-evaluation \
  --include 'wsm/scene-001/*' 'captions/scene-001/*' \
            'manifest.jsonl' 'metadata.jsonl' 'README.md' 'checksums.json'

Authentication is required while repository access is restricted. Downloaded metadata uses relative paths and contains no internal storage locations.

Downloads last month: 205

Collection including nvidia/Cosmos3-Driving-Simulation-Evaluation

Cosmos3-Datasets

Collection

12 items • Updated 1 day ago • 1