This is a dataset of synthetic images of virtual scenes.

The dataset consists of 9,400 multi-viewpoint photoreal frames from a virtual city, each with a resolution of 1,280×960 and 13 categories of pixel-level semantic annotations, namely mixed scenes, sky, buildings, roads, sidewalks, fences, vegetation, poles, cars, signs, pedestrians, cyclists, and lane markings.

This dataset can be used to solve the problems of semantic segmentation and scene understanding of driving scenes.

A subset of this dataset:

• SYNTHIA-AL, which is a subset of video streams generated at 25 FPS and contains a lot of ground truth information.
• SYNTHIA-SF, which is a subset of video sequences generated at 5 FPS, contains 6 sequences.
• SYNTHIA-RAND, which is a subset of the original 13,407 images.
• SYNTHIA-RAND-CITYSCAPES, a subset of 9,000 random images.
• SYNTHIA VIDEO SEQUENCES, which is a subset of videos generated at 5 FPS and contains 7 sequences.

Example image: