UrbanSyn Synthetic Driving Scene Dataset

UrbanSyn provides over 7,500 photorealistic synthetic images of driving scenes. The dataset was created by the UrbanSyn organization to address the synth-to-real domain gap and includes ground-truth annotations for multiple vision tasks. It was last updated on the Hugging Face platform in April 2025.

Use Cases

• Training semantic segmentation models using the provided ground-truth semantic segmentation labels.
• Developing monocular depth estimation algorithms with the included scene depth annotations.
• Training and evaluating object detection models using the 2-D bounding box annotations.
• Researching panoptic segmentation with the available panoptic instance segmentation labels.
• Benchmarking domain adaptation techniques across the synth-to-real gap using this synthetic dataset.

Strengths

• Contains over 7,500 annotated images.
• Provides four distinct types of ground-truth annotations per image: semantic segmentation, depth, panoptic segmentation, and bounding boxes.

Limitations

• The dataset size is relatively compact at 7.5k images compared to other large-scale autonomous driving datasets.
• All data is synthetic, which may introduce biases not present in real-world driving scenes.
• Specific details on class distribution, annotation quality metrics, or geographic diversity within the synthetic scenes are not provided.

Provenance

Source

UrbanSyn organization.

Collection Method

Synthetically generated photorealistic driving scenes.

Freshness

Last updated on the platform in April 2025.

Geography

Synthetic scenes; platform tags suggest a focus on US urban scenarios.