FROSI

Download of the full FROSI database.

The FROSI (Foggy ROad Sign Images) is a database of synthetic images easily usable to evaluate in a systematic way the performance of road sign detectors in foggy conditions.
The database contains a set of 504 original images 1400x600 with 1620 road signs (speed and stop signs, pedestrian crossing) placed at various range, with its ground truth.
The ground truth of each sign is made of its location, size and its distance to the camera.
Regulatory signs' distances range from 36 m to 300 m.
Accessing to the depth map of each image, it is possible to add an uniform fog relying on the Koschmieder model.
For each image, a set of 7 types of uniform fog were produced with visibility distances ranging from 50m (heavy fog) to 400m (slight fog).

Therefore FROSI set is made of a total of 3528 images (504×7) containing 11340 road signs (1620x7) placed at a distance less than 300 meters from the vehicle. All traffic signs have the same physical size, but when projected onto the image, they appear with a spatial extent inversely proportional to their range (from 16x16 to 200x200).

The authors of the database are:

• Dominique Gruyer
• Houssam Halmaoui
• Rachid Belaroussi
• Aurélien Cord