Spatio-Temporal Road Image Sequences For World Models

STRIDE contains approximately 82 billion tokens arranged into 6 million visual sequences derived from 131,000 panoramic road images. Developed by Tera-AI, this dataset combines imagery with metadata and highway system data to enable generative world model training. The dataset was last updated in August 2025.

Use Cases

• Train generative models to predict future road scenes using sequences of panoramic images.
• Develop spatio-temporal reasoning models using the 6 million visual token sequences.
• Analyze road infrastructure and highway systems by linking panoramic images with open highway data.
• Benchmark computer vision models on large-scale, sequential panoramic imagery from 131k source images.

Strengths

• Contains approximately 82 billion tokens for large-scale model training.
• Generates 6 million visual sequences from 131,000 source images, a 27x information increase.

Limitations

• Dataset is derived from a relatively small set of 131k base images, which may limit scene diversity.
• Specific row count, sample data, and file formats are not provided for evaluation.

Provenance

Source

Tera-AI via Hugging Face.

Collection Method

Sequences generated from panoramic images combined with metadata and openly available highway system data.

Freshness

Last updated in August 2025.